Frame Structural Analysis using the Stiffness Matrix

Method in Microsoft Excel and ETABS

CE 155 Matrix Structural Analysis Project

2nd Semester AY 2014-2015

Submitted by:
Carl Chester Ragudo

Submitted to:
Prof. Oscar Victor Antonio

I.

INTRODUCTION
A planar structural element such as a beam or column element has six
unknowns: two forces and a moment on each side. The total number of
unknowns of a planar structure would then be equal to thrice the number of
nodes in the structure. When 3-dimensional frames are considered, each
node would present a total of 6 unknowns: forces and moments on each of
the 3 axes. As the complexity of the structure increases, the computational
effort to determine all the unknowns increases exponentially. Matrix structural

analysis is now introduced to address this concern. Below is the given

structure for this analysis project.
Figure 1. Free body diagram of the structure
The structure consists of 20 nodes labeled A to Q with 31 members
using the nodes as joints as shown in Figure 1. With 6 degrees of freedom for
each node, a total of 120 degrees of freedom must be analyzed. Nodes A, D,
H, K, N, and R are defined as fixed supports. These restraints make the
displacements of the nodes equal to zero. The analysis is now left with 84
displacements and 36 support reactions to solve for. A manual computation
was done to solve for these unknowns using the stiffness matrix method
using Microsoft Excel. Another analysis was made using ETABS to compare
the results from the manual computation with.
II.

METHODOLOGY
a. Stiffness Matrix Method
The goal of this method is to generate the global stiffness matrix which
relates the displacements of each node to the forces applied on each node.
The global stiffness matrix is assembled from the element stiffness matrix

generated from each of the members using the matrix below. The matrix is
dependent on the following parameters: length of the member (L), crosssectional area (A), modulus of elasticity (E), modulus of rigidity (G), Poissons
Ratio (), polar moment of inertia (J), and moments of inertia about the z and
y axes (Iz and Iy). Information about the section properties needed were
obtained from an online source.

K =

The element matrices were then generated using the material and
section properties for each member. The generated matrices now relate the
displacements on their end nodes to the forces applied on them on its local
coordinates. To transform the element matrix to the same coordinate system
as the global stiffness matrix, a Transformation Matrix was used. The
transformation matrix relates the local forces to its corresponding equivalent
in the global coordinate system. It is constructed using the relative angles of

the local coordinate system of the member with the global coordinate
system. The definitions of the terms in the matrix are summarized in the
table that follows.
Table 1. Relative angles of coordinate axes
Angle
x

definition
Global x to local
x

Global x to local
y
z
Global x to local
z
x
Global y to local
x
y
Global y to local
y
z
Global y to local
z
x
Global z to local
x
y
Global z to local
y
z
Global z to local
z
The transformation matrix also applies to displacements thus {} =

[] {} and {F} = [K] {}. Substituting {} will yield

{ F }=

as proven in

class. [K] now corresponds to all the factors multiplied to { } .

[ K ] =
To further simply the computations, the terms attributed to the free
nodes and support nodes could be segregated and the global stiffness matrix
could be rearranged. To skip this step, the members were arranged starting
from the members connected at the free nodes so that the nodes at the
support will already be separated. The labeling started with member QG as
member 1 and member GF as member 2 and so on. When the global stiffness
matrix was assembled, the sections for the supports are already segregated.
The global stiffness matrix is now arranged as follows:

{ }[

Pf
K
= ff

Ps
K sf

]{ }

K fs f
K ss s

Further expansion of the equations show that the applied loads could
be represented as fixed-end forces and the previous equation becomes:

{ }[

Pf
K
= ff
Ps
K sf

]{ } { }
F

P
K fs f
+ fF
K ss s Ps

Where the f subscript correspond to the free nodes and s for support.
Rearranging the equation, the displacements of the free nodes can be
computed using
1

{f }=[ K ff ] {P f P Ff }
And support reactions from

{ Ps }=[ K sf ] { f }+{ PFs }

Since there are no forces applied on the joints,

Pf

can already be

taken as zero. Likewise, there are no applied forces on the support nodes as
well thus

{P Fs } is also equal to zero.

b. ETABS
As a control, the same structure was also analyzed using a
commercially available software. Although it is a paid software, its trial
version contains enough features to analyze the structure as required.
From the start up interface of the program, a grid system was defined
on the x-y plane as dictated by the program. The project was defined to have
the ground level of the structure on the x-z plane, however, the environment
of the program constructs the structure on the x-y plane which it uses as well
for its plan view. Adjustments for this matter were made later for the
comparison of the results of the two methods.
The grid was set-up with 3 nodes along the x-axis at 3745.45 mm and
2 nodes along the y-axis at 2746.66 mm. Three stories were created with
ceiling elevations of 3500 mm, 7500 mm, and 11000 mm. For consistency, kN
and kN/mm were used as units for force and moments as mm were used for
lengths since displacements are generally on that scale. Three section
properties were used for the structure: W14x43, W16x26, and W18x35. The
easy interface of ETABS allows drawing of the members while picking its
appropriate section properties. All members were set use A36 steel. All
material and section properties were already pre-loaded on the program. The
support nodes were set as fixed supports and the loading cases were entered
using its guided dialogue box.
When all of the necessary information had been loaded on the
program, the analysis of the structure was run. ETABS displays it results as
nodal displacements and support reactions drawn on the model with an
option to display the deformed shape of the structure.
III.

RESULTS AND DISCUSSION

Calculations made for the direct stiffness method were consolidated in
an Excel file. Each element stiffness matrix was generated for the 31
members and assembled to form the global stiffness matrix. The distributed
loads were translated to fixed end forces with a separate sheet on the file for

their computations. The nodal forces and moments used for the analysis are
listed below in Table 2.
Table 2. Fixed-end forces
FxQ
FyQ
FzQ
MxQ
MyQ
MzQ
FxG
FyG
FzG
MxG
MyG
MzG
FxC
FyC
FzC

MxC
MyC

MzC

0.00
23.73
0.00
3789.75
0.00
9645.03
0.00
23.73
0.00
3789.75
0.00
9645.03
-19.07
23.73
0.00

3789.75
0.00
11547.5
1

FxF
FyF
FzF

0.00
2.04
0.00

FxP
FyP
FzP

0.00
2.04
0.00

FxE
FyE
FzE

0.00
0.00
0.00

FxO
FyO
FzO

0.00
0.00
0.00

MxF
MyF

394.87
0.00

MxP
MyP

-394.87
0.00

MxE
MyE

0.00
0.00

MxO
MyO

0.00
0.00

MzF
FxJ
FyJ
FzJ
MxJ
MyJ

734.27
0.00
2.04
0.00
394.87
0.00

MzP
FxT
FyT
FzT
MxT
MyT

734.27
0.00
2.04
0.00
-394.87
0.00

MzE
FxI
FyI
FzI
MxI
MyI

0.00
0.00
0.00
0.00
0.00
0.00

MzO
FxS
FyS
FzS
MxS
MyS

0.00
0.00
0.00
0.00
0.00
0.00

MzJ
FxM
FyM
FzM

MzT
FxB
FyB
FzB

-734.27
-41.21
0.00
0.00

MzI
FxL
FyL
FzL

0.00
0.00
0.00
0.00

MzS

0.00

MxM
MyM

-734.27
0.00
23.73
0.00
3789.7
5
0.00

MxB
MyB

0.00
0.00

MxL
MyL

0.00
0.00

MzM

9645.0
3

MzB

35320.
90

MzL

0.00

The forces in Table 2 together with the [K ff] portion of the global
stiffness matrix were used to compute for the displacements of the free
nodes. This is possible because the displacements at the supports are equal
to zero. Table 3 lists the displacements from this operation.
Table 3. Displacements of free nodes from Direct Stiffness Method
uQ

wQ

0.84624
0.15906
0.29747

wF

xQ
yQ

0.00009
0.00154

xF
yF

zQ
uG

0.00020
5.54723
0.12463
-

zF
uJ

vQ

vG
wG

uF
vF

vJ
wJ

5.47708
0.07885
0.00646
0.00002
0.00103
0.00018
5.45873
0.07108
-

uP
vP
wP
xP
yP
zP
uT
vT
wT

0.82420
0.11172
0.00689
0.00001
0.00102
0.00004
0.82141
0.03394
-

uE

uO

vE

3.03364
0.03531

vO

0.31137
0.05042

wE

0.05295

wO

0.05294

xE
yE

0.00000
0.00057
0.00036
2.99335
0.05059
-

xO
yO

0.00000
0.00057
0.00006
0.31360
0.02158
-

zE
uI
vI
wI

zO
uS
vS
wS

xG
yG
zG
uC

0.29792
0.00012
0.00148

xJ
yJ

xC
yC

0.00020
5.53739
0.08734
1.98727
0.00009
0.00122

xM
yM

zC

0.00000

zM

vC
wC

zJ
uM
vM
wM

1.98338
0.00001
0.00122
0.00021
0.81990
0.09255
1.98755
0.00013
0.00126
0.00021

1.98335
xT
yT
zT
uB
vB
wB
xB
yB
zB

0.00001
0.00123
0.00003
3.14721
0.02484
0.96392

xI
yI

0.00003
0.00081
0.00114

xL
yL

zI
uL
vL
wL

zL

1.01677
0.00002
0.00070
0.00060
0.30522
0.03915
0.96360

xS
yS
zS

1.01681
0.00002
0.00070
0.00008

0.00002
0.00081
0.00008

Displacements are represented in millimeters and rotations are in

radians. The obtained displacements was then used to compute for the
support reactions using the [K sf] portion of the global stiffness matrix. Table 4
lists the obtained support reactions of the structure.
Table 4. Support Reactions from Direct Stiffness Method

FxA
FyA

FzA

MxA

MyA
MzA

FxD
FyD

FzD

MxD

11.496
2
11.536
97
0.9559
1
1713.0
8
8.0292
2
31719.
83
23.929
2
16.402
81
0.0638
1
108.86
3

FyH

19.314
5
23.500
04

FzH

1.0306
21

FzN

MxH

1823.7
59

MxN

FxH

MyH
MzH

FxK
FyK

FzK

MxK

FxN
FyN

2.1356
5
23.421
66

-6.9618
39945.
92

MyN

-1.7115
18.186
56

FxR
FyR

-0.0646
108.21
4
5.6053
7
4302.8
34
1.6940
2
10.024
8

-0.9749
1729.7
7

FzR

1.0286
06

MxR

1824.1
95

MzN

MyD
MzD

5.6239
7
45564.
17

MyK
MzK

8.0291
3
3772.3
62

MyR
MzR

6.9624
8
3800.4
36

The support reactions are in terms of kN and kN/mm for forces and
moments respectively. The maximum displacements are recorded at 5.5 mm
and as low as 0.006 mm. rotations are fairly unnoticeable ranging around
0.0008 rad. As for support reactions, majority of the forces are on the positive
y-axis resisting the download distributed load while the forces on the x-axis
resist the horizontal point load and horizontal component of the distributed
load on the inclined member.
Since ETABS is an integrated program, results of the analysis was spot
on. Table 5 and 6 lists the results of the analysis from ETABS.
Table 5. Displacements from ETABS
uQ
vQ
wQ
x
Q
y
Q
z
Q
uG
vG
wG
x
G
y
G
z
G
uC

6.2
-0.2
-0.3
0.00013
0.00168
1
0.00021
7
1
-0.2
-0.3
0.00010
4

uF
vF
wF

6.1
-0.1
0.02211

uP
vP
wP

1
-0.2
0.02168

uE
vE
wE

xF

xP

zF
uJ
vJ
wJ

-0.0002
6.1
-0.1
-2.2

zP
uT
vT
wT

xJ

-2.7E-05
0.00136
5
0.00024
1

xT

-5E-06
0.00113
1
0.00005
1
-0.1
-2.2
0.00001
3
0.00136
6

xE

yF

-2.8E-05
0.00113
7

0.00176
0.00021
8
6.2

yJ

zI
uL

-0.1
2.2
0.00014
1
0.00143
3
0.00025

vB
wB

-2.6E-05
3.5
0.04232
1.1
0.00003
6

vC
wC
x
C
y
C
z
C

-0.1
2.2
-9.4E05
0.00138
6
-3.1E05

vM
wM
x
M
y
M

zJ
uM

zM

yP

yT
zT
uB

xB
yB
zB

0.0009
0.00122

yE
zE
uI
vI
wI
xI
yI

vL
wL
xL
yL
zL

3.4
-0.1
0.1
-1.1E05
0.00063
3
0.00041
3.4
-0.1
-1.1
-4.2E05
0.00078
8
0.00065
0.4
-0.1
1.1
0.00004
1
0.0009
-9.6E05

Table 6. Support Reactions from ETABS

uO
vO
wO
x
O
y
O
z
O
uS
vS
wS
x
S
y
S
zS

0.4
-0.1
0.1
0.00000
4
0.00063
3
-0.00067
0.4
-0.03659
-1.1
-2.8E-05
0.00078
8
-0.00009

FxA
FyA
FzA

MxA
MyA
MzA

FxD
FyD
FzD
MxD
MyD
MzD

11.945
3
20.749
8
-1.0511
1878.1
3
-8.64
33342.
5
23.180
1
29.312
3
-0.091
-147.41
-6.08
44702.
3

FxH

19.2226

FxN

FyH
FzH

30.3184
1.1101

FyN
FzN

MxH
MyH

MxN
MyN

MzH

1988.23
-7.75
40242.7
3

FxK

-1.7622

FxR

FyK
FzK

FyR
FzR

MxK
MyK

26.6795
-1.0413
1866.53
-8.64

MxR
MyR

MzK

4060.75

MzR

MzN

-2.2504
35.612
2
-0.0631

-115.08
-6.07
4620.1
5

-1.9204
18.088
8
1.1365
2018.9
3
-7.57
4281.2
8

To compare the results of the two methods, the relative deviation was
computed. In general, there is no conclusive relationship between the results
of the two methods. Although there are results that with deviation almost
equal to zero, there are those that have more than 50% deviation.
Interestingly enough, although the deviation of the displacements have a
wide range, the deviation of the results of the manual computation for the
support reactions have relatively closer results with deviations of about 3-5%.
ETABS has a feature to display the free body diagram of the structure
together with the computed support reactions. As shown in Figure 2 and 3,
the support reactions for each of the support joints are presented as directed
arrows in the joints. Note that the drawn structure on the program had a
different coordinate orientation though it is still consistent with the given
structure. To clearly visualize the results, the deformed shape of the structure
can also be generated. Since the displacements of the joints are insignificant
with respect to the actual lengths of the members, the deformed shape is
exaggerated to visualize the relative magnitudes of the displacements with
respect to each other. Figure 4 shows the deformed shape of the structure.

Figure 2. Support Reactions (Forces)

Figure 3. Support Reactions (Moments)

Figure 4. Deformed shape of the structure

Several factors may contribute to the difference in the answers
obtained from the two methods. The most probable error can be attributed to
clerical error. Because of the scale of the amount of information to be
processed, there might have been a tiny error in the assembly of the global
stiffness matrix. Although it can be assured that the procedure has been
thoroughly checked, some errors that can be attributed to data handling,
such as rounding off and values of Excel can produce small errors that can
cause the results to greatly deviate.
IV.

CONCLUSION

This project aimed to demonstrate the process of how matrix structural

analysis is used to compute for the large amount of unknowns normally
encountered in more complex structures. Although the basic concepts of
analysis can be done easily with pen and paper, its real life application would
require more computing power as the difficulty of the analysis grows
exponentially with increasing number of elements in the structure.
Using Microsoft Excel, it has been shown that manual computations for
matrix analysis using the direct stiffness method can be done. Excel can
easily reproduce repetitive calculations such as the computation of the
element stiffness matrices that must be done for all members of the
structure. Although not entirely automated, this project has shown the great
flexibility of Excel in processing vast amounts of data all the while motivating
the user to come up with ways to optimize how the computations can be
handled thus little by little automatizing the procedure to just entering the
bare inputs. But because of the scale of the computations, it should be noted
that analysis requires great care as small errors in the procedure can produce
big deviation from desired quantities.
This also has been a great opportunity to be familiar with structural
software such as ETABS. The integration of the analysis and the design
components of the program can be useful. The analysis it can make can be
made as close as possible to reality because of the versatility it offers with
the attributes and properties of materials already encoded in the program.

References
McGuire, W., Gallagher, R. H., & Zeimiam, R. D. (n.d.). Matrix Structural Analysis.