Вы находитесь на странице: 1из 32

CIVE 701

FINAL ASSIGNMENT

DUE THURSDAY, DECEMBER 14TH, 2017

MARIA RAGGOUSIS

DREXEL UNIVERSITY
FALL 2017
Maria Raggousis CIVE 701 1 Drexel University 12/14/2017

Table of Contents
Problem Statement ................................................................................................. 2
Executive Summary ............................................................................................... 3
I.) Flexibility & Stiffness ...................................................................................... 4
1.) 9 Degrees of Freedom ................................................................................. 5
2.) 5 Degrees of Freedom ............................................................................... 11
i.) Lateral Displacements at b & d .............................................................. 12
ii.) Lateral and Vertical Displacement at c .................................................. 13
3.) Comparison ............................................................................................... 14
4.) Conclusion ................................................................................................ 18
II.) Design Envelope ............................................................................................ 19
1.) Influence Lines.......................................................................................... 21
2.) Design Envelope ....................................................................................... 26
3.) Member Design ......................................................................................... 29
4.) Conclusion ................................................................................................ 31
Maria Raggousis CIVE 701 2 Drexel University 12/14/2017

Problem Statement
Maria Raggousis CIVE 701 3 Drexel University 12/14/2017

Executive Summary
The gable frame as illustrated in the problem statement will be analyzed with two main purposes. The first is
to develop and understand the flexibility and stiffness matrices of a system and compare results obtained by
such a method with slope-deflection methods and a computer finite element model analysis. The second is
to utilize this understanding and develop a maximum design envelope for a set of forces on the same frame.

Flexibility and stiffness are a function of the section properties of a system and the structure’s geometry. For
a given section property, there is only one unique stiffness matrix and flexibility matrix for the structure;
however, there are infinite stiffness and flexibility matrices available to describe the same geometry. This is
why it is important to specify and note what modulus of elasticity, moment of inertia, and cross-sectional
area is used for a system. The flexibility and stiffness matrices are symmetrical and they are inverses of each
other. As flexibility increases, stiffness decreases and vice versa. The matrices are a way to represent the
linear system of equations that relate the forces placed on the structure at the joints (fixed end forces) to the
displacements resulted from each degree of freedom. A structure has many degrees of freedom but many of
them are constrained or related to each other. In this scenario, the structure has five independent degrees of
freedom and all other deflections can be obtained if any five are independently obtained. The flexibility and
stiffness matrix approach to structural systems is an exact approach the same way slope-deflection is and
they are both a function of equilibrium and deflection criteria. We expect and have obtained identical results
in both methods.

The second analysis objective in this assignment is to create a maximum demand envelope and design the
member accordingly to this demand envelope. Typically in introductory structural engineering concepts, one
static load is applied to the structure and we are observing it’s reactions to these loads in the form of internal
and external axial, shear, and bending moment forces. To design a member safely and efficiently, one must
obtain a worst-case scenario for the system based on all the possible forces that could be applied to it. The
goal here is to figure out which loads, when placed together, amplify each other and increase the total internal
forces experienced. Applying this to the entire structure becomes a time-consuming and tedious task. To
efficiently do this, we must first obtain the regions which will consistently experience the worst effects of the
loads. We are trying to find the absolute maximum effects of the loads on the structure and so if an area is
consistently experiencing the most reactions, there is a good chance it will also be the point of the maximum
design envelope. Typically, the critical locations are the columns of a structure since they are the only path
down to the ground, where all loads eventually must travel to.

The design envelope consists of the maximum positive and negative axial force and bending moment
experienced in the columns, particularly the base of the columns. Under normal loading conditions, the design
envelope will be a constant value for the system regardless of the section properties. This scenario is an
extraordinary circumstance because we are also applying physical displacements to the joints where no matter
what cross section it is, we must move this join one inch. Conceptually, trying to move a joint for something
that gets stiffer and stiffer must get more difficult and so the internal forces experienced for such a system
actually increase. Because of this, our design envelope increases with increasing structure stiffness. This
complicates the design of the system but through a iterative process, the design may be obtained where neither
axial or bending moment forces are in violation of the demand created from the loads.
Maria Raggousis CIVE 701 4 Drexel University 12/14/2017

I.)Flexibility & Stiffness


A powerful tool in analyzing structural systems is with member stiffness and flexibility. These two physical
phenomena are inverses of each other. This is intuitive: the higher the flexibility, the lower the stiffness and
vice versa. A system must be discretized into distinct nodes and connecting members. It is ideal to describe
a system with the least nodes as possible for simplicity. The nodes are typically in a change of geometry,
restraint, or boundary condition. For the gable frame as shown, there are five distinct nodes and four
members. Due to the fixed restraints at the ground, there will never be any displacement at these locations.
In this assignment we will be studying the three nodes at the top of the frame. Each node starts with three
degrees of freedom in an unconstrained model: displacement in the X, displacement in the Z, and rotation
about Y.

Stiffness and flexibility are a function of the actual member geometry, material properties, and length.
Because of this, we must explicitly state our analysis parameters. In order to more efficiently compare the
results from this assignment to that of assignment 3 (slope deflection methods of analysis) for the same frame,
the same parameters will be assigned.

The member properties in SAP2000 are:


𝐴 = 1 𝑓𝑡 2
1 4 Therefore:
𝐼= 𝑓𝑡
12 1
𝑘 2𝐸𝐼 1 ∗ 90 ∗ 12
𝐸 = 90 2 = =1
𝑓𝑡 𝐿 15

To be explicit, an elastic modulus of 90 ksf and a moment of inertia of 1/12 ft4 is a ridiculous statement and
extremely low. Since this task does not revolve around design of the members, the particular values we study
do not actually matter. This is to have ease of calculation as this stiffness coefficient becomes 1. Since the
previous assignment was done in this way, by assuming the same properties it is the simplest way to ensure
the methodology presented in this report is accurate – the values may be checked against Yuan-Yu Hsieh and
S.T. Mau’s Elementary Theory of Structures, Chapter 8, Section 8-7, pages 246-250.

A stiffness matrix has elements kij where kij is the reaction function at i due to a unit displacement at j
when all other displacements are constant.

A flexibility matrix has elements fij where fij is the displacement at i caused by a unit force j when no other
forces are present.

These matrices relate the forces applied to a structure (P) to the structures stiffness (K) or flexibility (f) to
produce the displacements and rotations (U) generated from said forces:

[𝑃] = [𝐾][𝑈]
[𝑓][𝑃] = [𝑈]
[𝐾] = [𝑓]−1 & [𝑓] = [𝐾]−1
Maria Raggousis CIVE 701 5 Drexel University 12/14/2017

1.) 9 Degrees of Freedom


For three free nodes (b, c, d) and three degrees of freedom for each (dispX, dispZ, rotY) there exists nine
total degrees of freedom that affect our total global displacements.

First, the flexibility matrix with axial deformation permitted is obtained by creating 9 separate models. On
each model, a unit force is applied in a different degree of freedom each time. The displacements and rotation
at nodes b, c, and d, were extracted from SAP2000 for each of the 9 cases, this results in a 9x9 matrix. The 9
models, their applied unit forces, and the corresponding deflected shape from that unit load is shown below.
Because of the order the functions are obtained and populated in the table, the system is automatically
symmetrical without any alterations. The flexibility and stiffness matrices must be symmetrical because any
applied function has a corresponding reaction function.

Applied FX Applied FZ Applied MY

1k

1k 1 k-ft
Node b

1k 1 k-ft

1k
Node c

1k 1 k-ft
Node d 1k
Maria Raggousis CIVE 701 6 Drexel University 12/14/2017

Following that procedure, the following 9x9 flexibility matrix is shown below, with the columns being the
applied force function and the rows being their corresponding displacement function.

This flexibility matrix is developed for a system with axial displacements permitted as noted from the matrix
where a vertical force on the column produces a vertical displacement. Another important note for this system
is that a unit force may be applied as any number – this changes their scale but not their relative magnitude
to each other. What this means is that if you apply a unit force of one million kips instead of one kip, all
values will be one million times higher. Simply dividing every value by one million returns back a matrix
that is equivalent had it been developed by applying one kip. This procedure was followed so that the results
had many more significant digits. Truncation errors are sensitive in matrix multiplication and so the more
significant digits that are retained throughout the process before arriving at a conclusion, the better the results
will be.
Maria Raggousis CIVE 701 7 Drexel University 12/14/2017

To generate a stiffness matrix, essentially the opposite must occur from the flexibility matrix – you do not
want your structure to displace at all except when you force it to. To accomplish this, one can assign fixed
restraints to every single node in the system. Then, an applied unit displacement must be created on only one
degree of freedom for one node. The reactions induced to that fixed support from the displacement at the
node is a measure of the stiffness of the system. This concept is illustrated in the figure below; all nodes are
restrained and the only displacement is that imposed manually by moving the entire node.

Applied ΔX Applied ΔZ Applied Y

1 rad
1 ft
1 ft
Node b

1 ft 1 rad

1 ft

Node c

1 rad
1 ft
1 ft
Node d

There are 3 nodes of concern and 3 degrees of freedom for each node. A total of 9 reaction values are obtained
for each of these 9 loading scenarios to determine the stiffness matrix.
Maria Raggousis CIVE 701 8 Drexel University 12/14/2017

Following the procedure, the following 9x9 stiffness matrix is shown below, with the columns being the
applied unit displacement and the rows being their corresponding reaction function.

One interesting thing to note about the stiffness matrix for this frame is that the bottom left and top right
corners contain zeros. This is because the system is fixed at all nodes and when imposing a displacement on
one column, the reactions are completely absorbed by the two closest fixed joints. The force effects never
arrive at the other column purely because of the infinitely rigid fixed connections.

In the same way the flexibility unit loads were increased for the purpose of significant digits, the stiffness
matrix was also generated for imposed displacements of one million feet and radians. This has no physical
meaning and is obviously outrageous – it’s purpose is to solely generate enough significant digits in the
reactions to have a truly accurate matrix. The entire matrix was then scaled down by a factor of one million
and compared to that obtained from applying a simple unit displacement of one foot. The results were
identical to 2 significant digits but the matrix which was scaled down had more significant power behind its
values.
Maria Raggousis CIVE 701 9 Drexel University 12/14/2017

It was previously stated that the flexibility matrix and stiffness matrix are inverses of each other. It is
important to always check if this is true before continuing with an analysis – if this is not true then there are
underlying method or software issues that must be addressed. One way to check for this is to take the inverse
of flexibility and compare each stiffness matrix value obtained from the inverse from that obtained by the
software reactions due to unit displacements. As illustrated below, one must compare if each value is equal
and by how much – this becomes tedious.

Does [𝑓]−1 = [𝐾] ?

𝑓1,1 𝑓1,2 ⋯ 𝑓1,𝑚 −1 𝑘1,1 𝑘1,2 ⋯ 𝑘1,𝑚


𝑓 𝑓2,2 … 𝑓2,𝑚 𝑘 𝑘2,2 … 𝑘2,𝑚
( 2,1 ) ( 2,1 )
⋮ ⋮ ⋱ ⋮ ⋮ ⋮ ⋱ ⋮
𝑓𝑛,1 𝑓𝑛,2 ⋯ 𝑓𝑛,𝑚 𝑘𝑛,1 𝑘𝑛,2 ⋯ 𝑘𝑛,𝑚

A shortcut for comparison is to multiply the two matrices together. Because they are the inverse of each
other, one should obtain the identity matrix with a value of 1 along the diagonal and zeros elsewhere.

When the flexibility and stiffness matrices were compared in this method when a unit force was used, the
identity matrix was not quite right – with some values that were supposed to be zero in the 0.1-0.5 range and
some values that were supposed to be 1 equal to 0.90-0.99. This is clearly close but it was not close enough
to satisfy that the identity matrix was obtained. By increasing the factors of displacements and forces to one
million and dividing again, a sufficient amount of significant digits are obtained. This allows the identity
matrix to truly become near identical.

The multiplication of the 9x9 flexibility matrix and 9x9 stiffness matrix including axial deformations is:

Denoted in green is the values that are equal to 1 or very nearly 1. In orange we see that there are a few values
that are essentially zero but not quite. Overall, this is an extreme improvement from previous iterations of
the matrix. The level of decimal points shown above would never be considered and so this is to illustrate
how close we actually are to having matrices be exactly the same, generated in two inversely related ways.
Maria Raggousis CIVE 701 10 Drexel University 12/14/2017

The next step is to repeat the procedure for a case in which axial deformations are not permitted. We should
note that while a flexibility may be obtained for all nine degrees of freedom without axial deformation, the
stiffness matrix corresponding to that is nonsensical and should not be considered. The stiffness matrix
without axial deformations should only be considered in the restricted models where we are not observing
any imposed axial displacements.

Following that procedure, the following 9x9 flexibility matrix without axial deformation is shown below,
with the columns being the applied force function and the rows being their corresponding displacement.

Thinking about this further, by increasing the area section property modification to be very large in
comparison, we essentially trick SAP2000 into not producing any displacements in the vertical direction of
the column. This means that not matter how much force is applied vertically at the column, it essentially does
not deform at all – the flexibility term for these functions is 0. When something does not displace no matter
how much force is applied, this suggests a stiffness that approaches infinity. The system must be infinitely
stiff in order to not displace at all which is why the stiffness matrix for such a phenomena does not truly
describe the system. Using the matrix rules, we prove this to ourselves by having entire rows and columns as
0 entries in the flexibility matrix and when the inverse of this matrix is taken, entries must be divided by
these rows and columns of 0 and 1/0 is infinity.
Maria Raggousis CIVE 701 11 Drexel University 12/14/2017

2.) 5 Degrees of Freedom


Although we have just developed stiffness and flexibility matrices for 9 different degrees of freedom, these
values are not independent of each other. To determine the independent degrees of freedom, we use the
formula:
𝐷𝑒𝑔𝑟𝑒𝑒𝑠 𝑜𝑓 𝐹𝑟𝑒𝑒𝑑𝑜𝑚 = 𝐺𝑒𝑛𝑒𝑟𝑎𝑙𝑖𝑧𝑒𝑑 𝐶𝑜𝑜𝑟𝑑𝑖𝑛𝑎𝑡𝑒𝑠 − 𝐶𝑜𝑛𝑠𝑡𝑟𝑎𝑖𝑛𝑡𝑠
= (3 𝑟𝑒𝑎𝑐𝑡𝑖𝑜𝑛𝑠) ∗ (5 𝑛𝑜𝑑𝑒𝑠) − 6 𝑓𝑖𝑥𝑒𝑑 𝑟𝑒𝑎𝑐𝑡𝑖𝑜𝑛𝑠 − 4 𝑟𝑖𝑔𝑖𝑑 𝑏𝑒𝑎𝑚𝑠
= 15 − 10 = 𝟓 𝑰𝒏𝒅𝒆𝒑𝒆𝒏𝒅𝒂𝒏𝒕 𝑫𝒆𝒈𝒓𝒆𝒆𝒔 𝒐𝒇 𝑭𝒓𝒆𝒆𝒅𝒐𝒎

Since we have 9 total free degrees of freedom, it becomes apparent that there are many different combinations
of degrees of freedom that can be selected as the independent variables. For this assignment, we will consider
two cases:
𝐶𝑎𝑠𝑒 1 → 𝜃𝑏 , 𝜃𝑐 , 𝜃𝑑 , ∆𝑏𝑥 , ∆𝑑𝑥
𝐶𝑎𝑠𝑒 2 → 𝜃𝑏 , 𝜃𝑐 , 𝜃𝑑 , ∆𝑐𝑥 , ∆𝑐𝑧

So how does one go about reducing the flexibility and stiffness matrices down to five degrees of freedom?

Because the flexibility matrix records displacements for unit forces, if we are only concerned about certain
degrees of freedom, we simply remove the entire rows and columns we are not interested in.

The stiffness matrix is not as straightforward. A new matrix must be obtained from the software with different
inputs than previously. The procedure is to constrain only the degrees of freedom that you are concerned with
and would like to appear in your end results. Apply the unit displacements to this modified frame and record
the reactions at the constraints that you have imposed. The reason to do this is that you do not want your
structure ‘wasting’ its reactions on areas you are not concerned with. When the entire structure is fixed all
over, a portion of the forces are dispersed in each reaction accordingly. We want to remove the structure’s
ability to carry the reaction for locations that are not independent. A new loading vector should also be
developed under these modified constraints. The restraints to obtain stiffness for each of the two cases is
shown below:

c c

Constrained Unconstrained

b d b d
Z

N/A Y X
𝐶𝑎𝑠𝑒 1 → 𝜃𝑏 , 𝜃𝑐 , 𝜃𝑑 , ∆𝑏𝑥 , ∆𝑑𝑥 𝐶𝑎𝑠𝑒 1 → 𝜃𝑏 , 𝜃𝑐 , 𝜃𝑑 , ∆𝑐𝑥 , ∆𝑐𝑧

All matrix work from this point forward is to be done without axial deformations permitted (no shortening
or lengthening of members).
Maria Raggousis CIVE 701 12 Drexel University 12/14/2017

i.) Lateral Displacements at b & d


The flexibility matrix for the system with observed lateral displacements at b and d is obtained by striking
out all rows and columns that are not included in our Case 1 degrees of freedom (𝜃𝑏 , 𝜃𝑐 , 𝜃𝑑 , ∆𝑏𝑥 , ∆𝑑𝑥 ).

The stiffness matrix is obtained from only the


constrained reactions (𝜃𝑏 , 𝜃𝑐 , 𝜃𝑑 , ∆𝑏𝑥 , ∆𝑑𝑥 ) from the
previously stated procedure and is shown as:

Multiplying the flexibility matrix and stiffness matrix together, we should obtain the identity matrix and we
essentially do. This confirms that the flexibility and stiffness matrix are accurate and inverses of each other.
Maria Raggousis CIVE 701 13 Drexel University 12/14/2017

ii.) Lateral and Vertical Displacement at c


The same procedure from section (i) should be followed. The flexibility matrix for the system with observed
lateral displacements at b and d is obtained by striking out all rows and columns that are not included in our
Case 2 degrees of freedom (𝜃𝑏 , 𝜃𝑐 , 𝜃𝑑 , ∆𝑐𝑥 , ∆𝑐𝑧 ).

The stiffness matrix is obtained from only the


constrained reactions (𝜃𝑏 , 𝜃𝑐 , 𝜃𝑑 , ∆𝑐𝑥 , ∆𝑐𝑧 ) from the
previously stated procedure and is shown as:

Multiplying the flexibility matrix and stiffness matrix together, we should obtain the identity matrix and we
essentially do. This confirms that the flexibility and stiffness matrix are accurate and inverses of each other.
Maria Raggousis CIVE 701 14 Drexel University 12/14/2017

3.) Comparison

To properly assess if these stiffness and flexibility matrices accurately describe the system we have shown,
the frame must be loaded and we should see how the structure deflects under those loads. The stiffness is a
function of the member properties and so a higher inertia or elastic modulus will result in less deformation.
2𝐸𝐼
Seeing as our member has a stiffness factor of = 1, we should be anticipating very large deformations –
𝐿
unreasonable deformations in reality but they exist solely to check with the previous assignment in which we
calculate the deformations at the degrees of freedom for the frame using slope-deflection methods.

The first thing we need to obtain is the theoretical or expected deflection values both with and without axial
deformations. These values were obtained in the previous assignment utilizing SAP2000 software for the
gable frame under a 1 kip per foot projected gravity roof load. They are restated here:

[U] (global displacements)


Degree of Freedom With Axial Deformations Without Axial Deformations
∆𝑏𝑥 128.75 129.23

∆𝑏𝑧 2.056 0.00002

𝜃𝑏𝑦 -0.506 -0.460

∆𝑐𝑥 1.848 1.855

∆𝑐𝑧 98.977 95.532

𝜃𝑐𝑦 -11.072 -11.071

∆𝑑𝑥 -69.251 -69.794

∆𝑑𝑧 2.109 0.000002

𝜃𝑑𝑦 8.439 8.385

We can notice how the displacements in the z direction essentially become 0 when we increase the area
without changing anything else about the system to make it infinitely harder to stretch the members out. To
maintain equilibrium and the geometric constraints, all of the displacements must redistribute to
accommodate for that change.
Maria Raggousis CIVE 701 15 Drexel University 12/14/2017

The final component to consider is the force vector to correspond with all of our degrees of freedom. For the
9x9 stiffness matrix, we must conduct a fixed end force transformation of the loads. For the 5x5 stiffness
matrix, the restraints are no longer completely fixed as discussed previously, we only restrain the degrees of
freedom that we are interested in and so the force transformation of the load changes. For example, node b
no longer should be fixed, it should be restrained only in Translation in the X and rotation in the Y to calculate
the force transformation for the constrained degrees of freedom considering only rotations and lateral
displacements at b and d. The software was utilized to develop the force vectors but one may refer to the
previous for the hand calculations of such a system.

In short, the force transformation for the 9x9 matrix becomes:

In vector format the forces become:

Function 9 Degrees of Freedom Constrained ∆𝑏𝑥 & ∆𝑑𝑥 Constrained ∆𝑐𝑥 & ∆𝑐𝑦
𝐹𝑏𝑥 0 6 -
𝐹𝑏𝑧 4.5 - -
𝑀𝑏𝑦 -6.75 -6.75 -6.75
𝐹𝑐𝑥 0 - 0
𝐹𝑐𝑧 12.5 - 12.5
𝑀𝑐𝑦 -14.5833 -14.5833 -14.5833
𝐹𝑑𝑥 0 -6 -
𝐹𝑑𝑧 8 - -
𝑀𝑑𝑦 21.333 21.333 21.333
Maria Raggousis CIVE 701 16 Drexel University 12/14/2017

We are now equipped to solve all of our flexibility and stiffness matrices for the displacement vector and
compare to the results for the [U] vector from the previous assignment (and restated on page 13).

One example will be illustrated to show the methods and all other matrix multiplication was conducted in the
same manner. The functions =MMULT(array1, array2) and =MINVERSE(array) were utilized in Excel but
the matrix multiplication may also be conducted in MATLAB.

Matrix multiplication is conducted by multiplying a row in the square matrix by the column vector and adding
up each term to obtain the a corresponding value of the results vector. This is how a system of linear equations
is arranged in a matrix that allows you to solve the equations simultaneously.

Taking the flexibility matrix for the lateral and vertical displacements at c as an example:

% Difference of [U]
P [U]
To Assignment 3

6 129.23 0.00003%

-6.75 -0.46 0.00075%

∗ -14.583
= -11.07 0.00000%

-6 -69.79 0.00000%

21.33
8.38 0.00006%

Dot Product

(0.147 ∗ 6) + (2.264 ∗ −6.75) + (−0.158 ∗ −14.5833) + (−.128 ∗ −6) + (0.445 ∗ 21.33) = 8.38

This matrix multiplication of [f][P] to obtain a calculated [U] from flexibility and [K]-1[P] to obtain a
calculated [U] from stiffness was executed for every matrix shown in the report thus far. Each of these
calculated [U] vectors was compared to the [U] vector that SAP2000 outputs as the total global displacements
of the nodes with the roof live load placed on top of it. The power of flexibility and stiffness matrices, once
obtained and confirmed that they are accurate, is that one may load the frame with any P vector at all and
instantly know how the system will deform, quantitatively. Or, one may wish to have a critical deformation
and be interested to know exactly the value of what forces will create these deformations.
Maria Raggousis CIVE 701 17 Drexel University 12/14/2017

This procedure is thus followed and the displacement vectors calculated for each matrix. The results are
tabulated:

9x9 flexibility 9x9 stiffness 9x9 flexibility 5x5 flexibility 5x5 stiffness 5x5 flexibility 5x5 stiffness
with axial with axial without axial ∆𝒃𝒙 & ∆𝒃𝒛 ∆𝒃 𝒙 & ∆𝒃 𝒛 ∆𝒄𝒙 & ∆𝒄𝒛 ∆𝒄𝒙 & ∆𝒄𝒛
∆𝒃 𝒙 0.00009% 0.00006% 0.00008% 0.00003% 0.00002% - -

∆𝒃 𝒛 0.00008% 0.00007% 0.00424% - - - -


𝜽𝒃𝒚 0.00043% 0.00005% 0.00036% 0.00075% 00.00085% 0.00036% 0.00036%

∆𝒄 𝒙 0.00422% 0.00188% 0.00428% - - 0.00425% 0.00431%

∆𝒄 𝒛 0.00000% 0.00000% 0.00000% - - 0.00000% 0.00000%


𝜽𝒄𝒚 0.00001% 0.00000% 0.00005% 0.00000% 0.00000% 0.00005% 0.00005%

∆ 𝒅𝒙 0.00012% 0.00006% 0.00011% 0.00000% 0.00001% - -

∆ 𝒅𝒛 0.00008% 0.00009% 0.01186% - - - -


𝜽𝒅𝒚 0.00023% 0.00020% 0.00024% 0.00006% 0.00006% 0.00024% 0.00024%

Clearly, the level of decimal points shown in this table is excessive by any standards but it’s purpose is to
show just how close the stiffness and flexibility models are able to predict the deflected shapes the same way
the finite element model does. Had the percentages been shown with only one or two significant digits, all of
the entries would be 0% error and while that is essentially true, we must show that there are some differences,
albeit negligible.

The largest difference in the table above comes from the displacement in the x direction at node c, consistently
this has the highest difference. Perhaps it has something to do with the fact that the c node is at the top of the
frame and in the middle of two systems and by assuming no axial deformations, the system must pull on the
frame on each side until reaching a deformed shape at equilibrium. There may be some effects in the finite
element model that are being described that the flexibility method is lacking. Another interesting note to
make is that the percent differences for the 5x5 matrices using the same degrees of freedom is almost
identical. This makes sense since the matrices are generated on the same system just in an inverse way. This
is another indication that they are almost perfect inverses of each other. From the identity matrix we see that
it is not a true identity matrix but it gets so close that it might as well be. Had we observed the percent error
from the flexibility matrix and its own inverse, the [U] vector output would have been the same and so the
percent errors would have been the same.
Maria Raggousis CIVE 701 18 Drexel University 12/14/2017

4.) Conclusion
So what can we conclude from this exercise? Many things have been learned in the process of developing
these matrices. Firstly, the flexibility matrix and stiffness matrix intuitively are inverses of each other (the
more flexible something, the easier it is to bend, the less stiff it becomes). Not only that, but they are obtained
in inverse ways – one is observing displacements from a force and the other is observing a force due to
displacements. When the two systems are exactly inverses of each other, the method has been executed
correctly.

It was also observed that when a small unit force or unit displacement was imposed (1 kip, 1 foot, 1 radian,
1 kip-foot), the stiffness and flexibility matrix were not producing an identity matrix very accurately. On top
of that, the errors of the [U] vector were large, between 1-20%. This did not seem appropriate seeing as this
method is an exact method to calculate displacements. The unit forces were increased by a factor of one
million to obtain the matrices and then the entire matrix was normalized back down to the equivalent of the
unit force. This produced many more significant digits as the system could respond to this greater load. What
was observed was that now the flexibility and stiffness matrices were almost exactly inverse of each other
and every single error term was reduced to something negligible. When dealing with matrix multiplication,
even small truncation produces large effects.

We can also both intuitively and numerically determine that although it would be nice to ignore axial
deformations in this system, when dealing with the global system it changes the physical meaning of the
system. If we say that the system will not deform axial no matter how hard we push on it, then it’s stiffness
is infinite. To combat this, we must constrain our degrees of freedom to only the independent ones and not
be concerned with the axial deformations (vertical displacements at b and d). Since these degrees of freedom
are not in our system of concern, then the stiffness matrix can be generated normally.

Finally, we can conclude that the flexibility and stiffness matrix method produces the exact same results as
the software when both the method and the software are utilized appropriately. This allows us to predict a
system’s deformation behavior under any loads just by developing the stiffness or flexibility matrix of the
system. It is important to note that there are infinite flexibility and stiffness matrices available for the
geometry of the system (the nodes, the lengths, & the fixities) but there is only one unique set of flexibility
and stiffness matrices for a particular section geometry. As soon as the inertia or section modulus changes,
the entire matrices will change. The matrix scales proportionately to itself, the stiffer the members, the lower
the values in the stiffness matrix become and utilizing the same force vector, a lower deformation vector will
be the result.

The flexibility and stiffness matrix method of analysis is an exact method and by obtaining the same results
as the software and slope deflection hand calculations, we confirm the validity of the method.
Maria Raggousis CIVE 701 19 Drexel University 12/14/2017

II.) Design Envelope


We wish to design this frame for additional loads than just the roof live load that was previously discussed.
The possible loads that could be present on the frame are shown below:

All of the loads shown above have options associated with them. Our task is to determine which particular
combination of each of these loads will result in the absolute maximum internal forces felt by the entire
structure. There are two major steps in determining this; first, we must determine at what location the
maximum positive and negative axial and moment forces are likely to occur – this is known as the critical
location, second, we must determine which combination of load will linearly add together to produce the
most positive and negative axial and moment forces at those critical locations.

To accomplish this, all options must be considered for each load. The roof live load may be present on the
whole frame, only the left portion, only the right portion, or not there at all. The wind is a distributed load
and can be applied in any direction on each of the members. The crane load may be anywhere along the span
but the closest it approaches is 1 foot from the column. Additionally, each column base may experience an
imposed displacement or rotation in any direction.
Maria Raggousis CIVE 701 20 Drexel University 12/14/2017

To determine what is the critical locations of this frame, we can apply all of these loads individual and observe
the axial force and bending moment diagrams. What we are looking for is the location of maximum internal
force and noting it. Below is a sampling of some qualitative moment diagrams where the arrows are indicating
a relative maximum axial or bending moment internal force in the frame.

From completing this exercise for all of the loads individually, it was determined that overall the columns
will experience the maximum forces in the structure. Specifically, the critical location is determined to be the
bottom of the column. Internal axial forces due to induvial loads tend to be constant in the column and so for
the axial forces, it doesn’t matter whether we specify the top of the bottom of the column. Looking at all of
the bending moment diagrams which are not constant throughout the column, it is determined that the
maximum moment in the structure will occur specifically at the base of the column.

From here we are now prepared to qualitatively determine what exact combination of loads will create the
worst effect on the base of the columns.
Maria Raggousis CIVE 701 21 Drexel University 12/14/2017

1.) Influence Lines

Since the system has loads with options, qualitative influence lines should be created. An influence line is
the way a function response changes as a load moves across a structure. Conceptually, we may use the crane
load as an example. The crane may exist anywhere between the joints B and D. Depending on where the load
is, the axial and moment force reactions at the joints will increase or decrease. Below is the qualitative
influence lines for various responses as the crane load moves horizontally between the columns:

As mentioned previously, the crane only approaches as close as one foot to the columns. The maximum axial
force therefore occurs when the load is one foot from the respective column and the maximum moment force
occurs when the load is one third of the length away from the respective column. Therefore only these four
scenarios are to be considered in the loading combinations since they produce the maximum responses.
Maria Raggousis CIVE 701 22 Drexel University 12/14/2017

A shortcut to obtaining these and any other influence lines is the Muller-Breslau principle. The Muller-
Breslau principle states that the influence line of a function in a structure takes the form of the deflected shape
when the ability to resist that same function has been removed and a unit displacement imposed in the same
direction. In the case of the crane, we are interested in the influence line for the vertical reaction at the end,
so we remove the beam’s ability to resist vertical movement at B and then we move it a unit displacement in
the vertical direction. The displaced shape becomes a triangle and the influence line is exactly equal to the
deflection at any point once it has been normalized to a unit load.

While there are shortcuts to obtaining numerical influence coefficients from Muller-Breslau’s principle such
as the conjugate beam method, we are only interested in the qualitative shape so that we know where to apply
the loads in the design analysis. Muller-Breslau’s principle is especially convenient for these statically
indeterminant and redundant systems without the need of actual values.

Betti’s Law, also known as the Maxwell-Betti reciprocal work theorem, works in conjunction with Muller-
Breslau, stiffness, and flexibility. It states that the work done by one force system over the displacements
caused by another force system is equal to the work done by the second force system over the displacements
caused by the first, or more simply: 𝑃1 ∆1,2 = 𝑃2 ∆2,1. This explains the symmetry in the flexibility matrix more
explicitly. In the flexibility matrix, we have certain displacements caused by a force but when we apply that
same force to the location of the first displacement, that same displacement is generated at the location of the
first force. Betti’s Law is particularly true in cases with multiple indeterminacies – take this as an example:
Maria Raggousis CIVE 701 23 Drexel University 12/14/2017

Now that the methods have been defined, they may be used. Since we identified the critical locations as the
base of the columns and we are interested in the maximum positive and negative axial force and bending
moment at this location, we must remove the structures ability to withstand these functions at that location
of interest and apply a unit displacement. From this, the deflected shape is the qualitative influence line.

We would like to obtain influence lines in order to understand in which direction the wind will create the
worst scenario for each of the four cases identified (axial and moment, + and -). The wind force which acts
normal to the members shall be placed in the direction of the influence line. The integral of the area under
the curve (the influence line) sums the affects of a distributed load placed on the length of the member. We
are not so concerned with the numerical computation of the area under the curve since the software will
compute it for us, all we must obtain is which direction to place the wind force in, the results of this analysis
are shown below.

Now that we understand which directions to apply the wind in order to see the worst effects on the structure,
we can do the same to the rest of the loads in question. Originally, it was thought that the Muller-Breslau
principle would apply to the roof live load as well; however, upon inspection it was found that the roof live
load did not make matters worse, it was actually negating some effects in some cases. This must be due to
the fact that the wind live loads are acting normal to the structure and the influence lines are for effects that
act normal to the structure as well. The roof live load is a gravity projected load and so some of the force
disperses into the axial forces of the beam, it is not purely perpendicular to the member.

This leads us to the second approach to determine which loads to place where: applying it as a unit load and
observing the direction and relative magnitude of the effect that force or displacement has on the structure’s
internal axial forces and bending moments. The directions of the roof live load and all of the manually
imposed displacements were developed in this manner.
Maria Raggousis CIVE 701 24 Drexel University 12/14/2017

From this qualitative analysis, it was determined that the roof live load only inflected negative moment at
node A, positive moment at node E, and negative axial forces in both columns A and E. The roof live load
was split into three scenarios, across the entire system, across the left half, or across the right half. The effects
from each of these were in the same directions as described above so one of them must be selected to cause
the maximum effect. The table below shows the values obtained, it is clear that we select the live load cross
the entire frame for maximum effects in -Mmax at A, +Mmax at E, -Nmax at A, and -Nmax at E. The roof
live load should not be present in all other loading combinations because it does not create those reactions.

BENDING MOMENT EFFECTS AXIAL FORCE EFFECTS


𝑨𝒃𝒂𝒔𝒆 𝑬𝒃𝒂𝒔𝒆 𝑨𝒃𝒂𝒔𝒆 𝑬𝒃𝒂𝒔𝒆
𝑹𝑳𝑳 → 𝑨𝒍𝒍 -25.9 21.9 -12.3 -12.6
𝑹𝑳𝑳 → 𝑳𝒆𝒇𝒕 -5.9 9.5 -7.5 -1.4
𝑹𝑳𝑳 -19.9 12.4 -4.8 -11.2
→ 𝑹𝒊𝒈𝒉𝒕

A similar approach was used for all of the displacements but those did not need values to be determined, their
effects are straightforward and opposite. We only need to apply each displacement in one direction for each
column. Since they are imposed displacements and act at the join of interest, a change in direction for the
displacement results in a change in direction for the internal forces. The effects of a displacement at one
column is still considered and included in the load combination for each of the columns to maximize the
effects on the structure. The qualitative effects from the positive direction of the displacements is shown
below. The negative effects are taken as the opposite of these diagrams.
Maria Raggousis CIVE 701 25 Drexel University 12/14/2017

From this, a final table can be created to illustrate exactly what load to place and in which direction. This
table is a set of instructions to load in to the SAP2000 load combinations. We must linearly add all of the
loads in the table to achieve the maximum positive and negative axial force and bending moment effects in
the system. The results of these loads applied together is the design envelope.
Maria Raggousis CIVE 701 26 Drexel University 12/14/2017

2.) Design Envelope

Finally, the load combinations as described in the previous table are created in SAP2000 by first creating
load patterns each containing only one of the rows in the table and then selecting which of these apply to
generate the maximum demand that you wish to obtain. Below is a sample of how the individual load patterns
and load combinations are created in the software.
Maria Raggousis CIVE 701 27 Drexel University 12/14/2017

The axial force (kips) and bending moments (kip-ft) are recorded for the base of the columns (nodes A and
E) for each of the load combinations to obtain the design envelope. It should be noted that normally for
external forces applied to a structure, assuming the structure is of uniform section, the section flexibility and
stiffness do not impact the internal axial force and bending moment diagram. That is to say that if we apply
1 kip to a simply supported beam at midspan, we will have 0.5 kip reactions at each end and the moment in
the center will be PL/4.

With that, one might say that if we change the section properties, we should obtain the same design envelope.
In this particular scenario, this is not the case. When external forces are applied to the system, the section
properties will not affect the values of the internal forces; however, when a displacement is applied to a
system, the stiffness and flexibility of the structure make a difference in how the structure feels these
displacements. A flexible structure will have no problem moving an inch and so its internal reactions will be
less than a very rigid system that one must impose great force to move it that same one inch. In our system,
we not only have displacements on our base joints, but we have six of them occurring simultaneously, one in
each degree of freedom on each fixed joint. Because of this, we should expect that our system generate
slightly different internal reactions if the section properties change. A very basic sensitivity analysis was
2𝐸𝐼
conducted in which a base model was studied with = 1 and a second model was studied with a W10x12
𝐿
section assigned to it.

The design envelope of these two different frames is tabulated below:

Axial Moment

2EI/L = 1 W10x12 2EI/L = 1 W10x12

+Nmax at A 4.32 4.62 28.4 17.53


-Nmax at A -25.96 -26.27 -54.63 -43.67
Joint A

+Mmax at A 2.98 2.65 62.71 73.69


-Mmax at A -18.99 -18.69 -98.94 -109.75
+Nmax at E 4.48 4.78 -27.61 -30.8
-Nmax at E -26.45 -26.76 49.21 52.32
Joint E

+Mmax at E -17.27 -17.02 96.63 106.29


-Mmax at E 0.88 0.63 -52.03 -61.88

2𝐸𝐼
In some cases, there is only a slight difference between the = 1 model with the W10x12 such as the axial
𝐿
forces in the +Nmax at A model of 0.3 kips but in the moment of the +Mmax at E model there is 10 kip-ft
difference. This difference depends on how important the displacements are to the overall reactions in the
system. In some cases, the rotation was producing negligible reactions but certain displacements were
magnifying the reaction results.
Maria Raggousis CIVE 701 28 Drexel University 12/14/2017

Finally, these design criteria can be plotted on a bending moment vs. axial force interaction plot:

This plot is visually showing us the maximum forced felt by the structure. On the x axis we have the bending
moment at the base and on the y axis the axial force at the base. What we see is that the columns are
experiencing forces almost exactly opposite of each other. Column A is denoted in blue and column E is
2𝐸𝐼
denoted in green. The square represents the system where = 1 and the triangle represents the W10x12
𝐿
model. The maximum bending moment occurs as approximately +100 k-ft in column E and -100 in column
A, nearly a mirror image. The axial forces in this column are not as symmetrical about the axis but the
columns are symmetrical to each other. The frame is not symmetrical in its geometry, but it is very close to
it. This agrees with the design envelope that the forces felt on the system are almost equal and opposite for
each column but not quite.

In more general terms, a design envelope is expressing the worst-case scenario for a structure and design
should be well outside these requirements. They gray shaded region represents an ‘envelope’ for the structure
as a whole. The darker shaded regions represent the more flexible section properties while the lighter region
which is just about slightly larger is for the stiffer system. What we see occurring is that the displacements
Maria Raggousis CIVE 701 29 Drexel University 12/14/2017

we imposed on the structure are having a harder time moving and so more force must be exerted internally
to accomplish the same movement. The design envelope increases slightly under these particular loading
conditions as stiffness increases. This is worth noting as we prepare for design because you may design the
section according to this design envelope only to find out that the section you have chosen is so stiff that the
design envelope has increased dramatically and the section chosen is not good enough anymore!

3.) Member Design

The member design will be conducted as a combination of software analysis and a rough estimate hand
analysis. First, the software will be instructed to select member sizes based on the load combinations we have
specified for maximum and minimum axial forces and bending moments for the base of the columns which
we previously decided were our critical locations. The cross sections that the software outputs will then be
checked with the AISC Manual Bending Moment Capacity charts (Tables 3-10). The design envelope for
these sections will be obtained and plotted. As mentioned before, the design envelope will change for every
different section we apply because an imposed displacement is a function of the member stiffness. If the
sections chosen by the software are acceptably on the outside of the region of the new design envelope, then
these sections will be chosen as the final design. Sections slightly more stiff and more flexible of those chosen
by the computer will be analyzed as well to see how their capacity compares with the envelope.

With the load combinations for maximum positive and negative axial force and bending moment as the design
criteria, the SAP2000 software designed the system as shown below. The approximate available moment
capacity is shown adjacent to each section label which was obtained from the AISC Manual as a function of
unbraces length.

34 𝑘 ∗ 𝑓𝑡 35 𝑘 ∗ 𝑓𝑡

105 𝑘 ∗ 𝑓𝑡 86 𝑘 ∗ 𝑓𝑡
Maria Raggousis CIVE 701 30 Drexel University 12/14/2017

In terms of the quality of this design, SAP2000 has successfully loaded members to approximately the same
range which means all members are feeling stress just about the same amount – there are no members that
are too stressed or members that are not feeling anything. Another design habit to observe is that the software
has successfully designed the columns to be larger sections than the gable frame. Since we established this
prior, the critical locations of this structure occur at the columns and so while they feel the most internal
forces, we also want to design them thicker than the beams or girders. The one glaring downside to this
exercise is that while column A appears to have been designed with a moment capacity greater than the
design envelope, column E has an available moment that is less than what we established was the demand!
Because of the imposed joint displacements, we must generate the demand envelope for these particular
sections in order to establish if the system is designed with enough capacity. Upon inspection of the moment
diagrams, these sections do not meet the available moment capacity so we must consider designing the
members without the auto selection of the software.

With the axial demands on the columns so low in relative magnitude, designing the member for axial forces
will almost never govern. For the sections we are using and unbraced lengths of 15-20 feet, the axial capacity
for these members is well over 200 kips while we expect to experience less than 50, typically. Therefore,
design for the moment capacity of the columns will be conducted. For simplicity and an increased factor of
safety, the same section shall be applied to the entire frame system, both the columns and the gable frame.
The initial design envelope was conducted on a frame with uniform section properties and so it is only fitting
to do the same in this analysis. Based on the estimated required moment that will be obtained, a section with
greater than 150 kip-feet capacity should be reasonable. A starting section of W12x40 (170 k-ft moment
capacity) for all members was used and the resulting design envelope obtained from SAP2000.

Based on the results, we find that the W12x40 has enough moment capacity for the maximum design
envelope of this system. The final design for the frame can now be selected as a W12x40 for all members.
Maria Raggousis CIVE 701 31 Drexel University 12/14/2017

4.) Conclusion
From completing this exercise, we now have a deeper understanding of what it means to find the worst
possible scenario for a structure. The design code would have simply requested us to use a slightly larger
factor against wind and the dead and live loads to create a combination that is typically the governing
scenario. In this case, we had many options for systems not considered in design codes such as the crane
loads. These loads which are optional or moving are best analyzed using influence lines. The location and
direction of the loads on the structure matters and we want to make sure that the absolute worst-case scenarios
are covered for any possible combination. We linearly add loading together only in the way to create more
internal forces – making sure not to put any system on it that would negate the effects, causing less of a
demand. If it turns out that wind only blows in one direction and the uplift is helping relieve some of the roof
live loads, then that is fine, our structure is experiencing way less stress than it was designed for. We must
design for these extreme cases, no matter how unlikely they are to happen, just in case they do happen.

Design of a structure comes with many more considerations than axial and bending moments. Deflections,
stability, and rare events are also limit states that must be considered in true design. An exercise such as this
allows us to practice one of these limit states of design which in this case was for strength. The governing
limit state must always be the true design, whichever that may be. We know that structures also have hidden
strength in them that is not specified in these codes and design tables. We absolutely want that hidden strength
because you never know when something will happen that is unexpected or not accounted for. We cannot be
designing systems so that they can take only their typical loading scenarios on an every day basis. We must
consider the long-term effects that this building will have to endure for a 50, 100, or 150-year service life.
Philadelphia alone is seeing structures survive well past a century and their designers would have never
thought the trauma that they might undergo.

Вам также может понравиться