Topology Optimization of Building Bracing Schemes

by

Zhen John Goo

B.S., University of Minnesota, 2011

Submitted to the Department of Civil and Environmental Engineering in Partial Fulfillment

of the Requirements for the Degree of

Master of Engineering in Civil and Environmental Engineering

at the
Massachusetts Institute of Technology

{MASS ACHUSETT7S .INS17UFE

June 2013 OF TECHNOLOGY

0 Massachusetts Institute of Technology. JUL 0 8 2013

All rights reserved.
LIBRARIES

Signature of Author:
Department of Civil and Environrontal Engineering
May 10 th, 2013

Certified by: Z
Jerome J. Connor
Professor of Civil and Environmental Engineering
Thesis Supervisor

Certified by:
Rory Clune
Massachusetts Institute of Technology
Thes s Reader

Accepted by: - I
Hidi Nepf
Chair, Departmental Committee on Graduate Students
2
 Topology Optimization of Building Bracing Schemes
by
Zhen John Goo

Submitted to the Department of Civil and Environmental Engineering in June, 2013,

in Partial Fulfillment of the Requirements for the Degree of Master of Engineering in
Civil and Environmental Engineering

Abstract

The thesis presents a technique for producing economical solutions for conventional
structural steel frames using topology optimization on the lateral bracing scheme. The
study focuses mainly on minimizing the amount of material used and optimizing the
placement of bracing elements in typical steel frame structures to achieve economical
and realistic solutions. Linear structural analysis is performed on steel frame
structures while considering static gravity and wind loading. The optimization scheme
uses a "multi-level design" approach with two distinct optimization loops. The optimal
beam and column sizes in a structural steel frame system are generated in the first
optimization loop and a bracing removal criterion is derived in the second loop to
optimize the lateral bracing topology. A space constraint is imposed on the steel frame
structure to enable designers to specify large empty spaces. A performance index is
proposed to compare the cost between structural steel frames designed using
conventional approaches, which rely on engineering experience and trial-and-error,
and the approach specified in this study, which uses a multi-step optimization scheme.
Two case studies are made, comparing steel frame structures designed using the
proposed method with one designed using the traditional method.

Keywords: Topology optimization, steel frame optimization, space-constrained

optimization, multilevel optimization

Thesis supervisor: Jerome J. Connor

Title: Professor of Civil and Environmental Engineering

Thesis reader: Rory Clune

3
4
Acknowledgements
I would like to express my deepest appreciation to all those who helped and guided me
for my time at MIT. These people made my thesis possible through support,
suggestions and encouragement. A special gratitude to my professor, Dr. Connor,
whose contribution in providing structural engineering knowledge and general wisdom,
helped me to be a better engineer and person.

Furthermore, I would also like to acknowledge my thesis advisor, Dr. Rory Clune, who
provided endless guidance with patience throughout the scope of my thesis, especially
in report writing. Special thanks goes my classmate, Zahraa Saiyed, who helped in
editing my report. Last but not least, I would like to thank all my fellow classmates in
the Master of Engineering program for being encouraging and thoughtful for the past
year.

5
6
Table of Contents
1.0 C hapter 1 - Introduction ................................................................................................. 11

1. 1 O u tlin e ..................................................................................................................---.----....------ 11

1.2 In tro du c tio n .........................................................................................................- -.--..------- 11

1 .3 S u m m a ry ................................................................................................................................. 13

1 .4 T h e sis O u tlin e ......................................................................................................... ..... ---.14

2.0 Chapter 2 - Literature Review ..................................................................................... 15

2 .1 O u tlin e ..............................................................................................................----- ------ .... 15

..------

2.2 Optimization Techniques in Structural Engineering.......................................15

3.1 C hapter 3 - M ethodology................................................................................................. 21

3 .2 O u tlin e ...........................................................................................--- --...--------- ---- ..................... 2 1

3 .3 C om putation al T ool........................................................................................................... 21

3.3.1 Optimization Algorithms - MATLAB ............................................................... 21

3.3.2 Structural Analysis Software - SAP2000 ...................................................... 22

3 .4 O ptim ization Sch em e ................................................................................................... 23

3.5 First O ptim ization Loop .............................................................................................. 25

3.6 Second O ptim ization Loop .......................................................................................... 31

3.6.1 Equivalent Frame Analysis for Reinforced Concrete Slab.....................32

3.6.2 Second Optimization Loop (continued) ........................................................ 34

4 .0 C h apter 4 - R esu lts ............................................................................................ .. -----------..... 41

4 .1 O u tlin e ................................................................... . ------ ..-- . 41

. - - - - -- - --.........................

4.2 Performance Index ............................................................................................. ... 41

4 .3 R e sults .....................................................................................--- . . . ---................... 42

Ca s e S tu d y I .............................................................. .... --.... . --------------------........................... 43

4 .3 .1 Case S tu dy II................................................................ ......... ----------------------.................. 45

5.0 Chapter 5 - Discussion................................................................... ...... 49

5 .1 O u tlin e ............................................................... .................... . - --- ---- - - - 49

----- - ---...........................--

5 .2 D iscu ssion ...................................................................--------.--.. . ----------------------...................... 49

5 .3 F u tu re W o rk ......................................................................... ..... ---------.............................. 50

6 .0 R eferen ces ................................................................................ . -------------------------....................... 51

A p p en d ix .........................................................--------- .... . . ------------------------............................................ 53

7
8
Table of Figures
Figure 1: Optimized topology layout of a 2D steel frame structure. (Liang 2000) 17

Figure 2: Left: Tripod designed using limited by the sphere volume constraint.
Right: A bridge created limiting all trusses beneath the deck. (Smith 2002) 18

Figure 3: Flowchart of general processes involved in the optimization scheme (detailed

process of first optimization loop and second optimization loop are shown in
Figures 3 and 5). 23

Figure 4: The first optimization loop process flowchart. 26

Figure 5: Section modifier user interface in SAP2000 required eight variables input. 27
Figure 6: The second optimization loop process flowchart. 31

Figure 7: The equivalent frame transformation that convert the stiffness in a concrete
slab to two cross-braced steel truss elements. 33

Figure 8: Process of gradually removing bracing elements. First, bracing removal to

obtain realistic bracing placement and then to satisfy space constraint. 36

Figure 9: Performance Index comparison in Case Study I. 45

Figure 10: Steel frame structure of two different methods in Case Study I. 45

Figure 11: Performance Index comparison in Case Study II. 48

Figure 12: Steel frame structure of two different methods in Case Study II. 48

9
10
1.0 Chapter 1 - Introduction
1.1 Outline
This chapter introduces the design of steel frame structures with and without
optimization. The chapter will discuss reasons that drive this study and briefly
introduce the content of the study. The last part of this chapter gives an outline of the
scope of this study.

1.2 Introduction
Steel has many desirable characteristics as one of the most common construction
materials in civil structure (McCormac 2011). Some of the properties of steel that
make it desirable for engineering are its malleability as well as high strength to mass
ratio. Steel is used abundantly as a construction material in many low- to medium-
rise buildings. Short steel frame structures are most commonly used as warehouses,
factories, and housings. These buildings mainly provide shelter and personal space for
individuals. Most of these buildings are built in large quantities, which make efficiency
of erection a very critical component of design. Experienced designers normally
generate conceptual designs for these structures from their experience. At times, this
knowledge may not be applicable to new projects and at other times, designers may
leave out important design factors. Most of the steel frame structures end up being
designed using trial-and-error, with the help of previous experience. The derivation of
economical lateral bracing schemes can be challenging due to a large number of
possibilities for the arrangement of bracings. A computational optimization scheme
can be used to derive economical conceptual designs of the steel frame structure is
investigated in this study.

The usage of optimization schemes in civil structure design helps designers to save
time, material and cost. Using optimization schemes, designers can avoid the
troublesome process of deriving designs through trial-and-error. Optimal solutions
should strive to be material and cost efficient. Application of optimization schemes to
civil structures is a tedious process as a large amount of iteration is required to be
computed before the solution converges. This is because realistic civil structures are
complex problems that are highly nonlinear (Connor 2003). Every optimization

11
iteration requires the structural response to be computed to progress to the following
iteration. When realistic civil structure models are used in optimization processes, the
computation of structural response takes a long time and hence the whole
optimization process becomes very time consuming. In the study, the structural
analysis part of the optimization scheme is done using a computational structural
analysis tool, SAP2000. In SAP2000, linear static frame analysis, one of the simplest
structural analysis methods, is used in the study to save computation time (Asif 2013).

In this study, short multistory steel frame structures are investigated. The
optimization scheme developed in this study adopts an approach that uses "multi-level
design" from the study done by Liang (2000). The approach involves two optimization
loops that optimize the gravity system of steel frame structures in the first run and
optimizes the topology the lateral bracing elements in the second run. The process of
deriving the conceptual design by having disintegrated gravity and lateral systems is
commonly used in conventional steel frame design (McCormac 2011). The technique
provides reasonable results and in some cases it gives a conservative design.

In the first part of the optimization scheme, the gravity system of the steel frame is
designed assuming continuous columns and simply-supported beams. Fixed-end
beams are rarely used in design due to the inefficient use of materials and costly
connection design (McCormac 2011). This study investigates the effect of the
placement of bracing elements, for which the objective cannot be achieved if moment
frames are used. All the bracing elements in steel frame structures are essentially
truss members that make up the building lateral system by acting in the axial
direction only. While designing these axially loaded bracing elements, only static wind
loading is taken into consideration. The assumptions for studying only static wind
loading are that seismic load is controlled by dampers and dynamic wind loading does
not control the design of short structures (Connor 2003). These assumptions simplify
the structural analysis method used in the study. Since the process of analyzing
structures takes up the largest portion of the time used for optimizing the design,
reducing the effective time of each structural analysis will drastically reduce the total
time of optimization.

12
An additional component is introduced to the optimization scheme. The new
component is a space constraint that defines the coordinates of null spaces in the
steel frame structure. These null spaces do not have any bracing elements running
through them. The null spaces can be defined as 2D planes or as 3D void spaces. The

2D planes can only be vertical or horizontal and the 3D spaces can only be in
rectangular boxes. However, designers can combine unlimited amounts of null spaces
to get desired space constraints. The purpose of these null spaces is to accommodate
large usable spaces or a more efficient usage of natural lighting. If a 2D null space is
placed on the exterior surface of the structure (same as the case studies in this
research), the structure can have unblocked window spaces to maximize the usage of
natural lighting. If a 3D null space is placed in the structure, the null area will be free
of all bracing elements and large usable spaces can be built.

One important feature of the study is that the steel frame structure is investigated in
3D space. This allows the investigation to capture more realistic structural responses
for designing the lateral system. A 2D model cannot investigate the third dimension
force effect, the coupling effect of forces in asymmetrically loaded structures (Connor

2003). When idealizing structures as a 2D model, the center of twist of a structure is

assumed to be aligned with the force. This is only true for a structure with a
symmetrical plan. When a structure does not have a symmetrical plan, the center of
twist will be shifted according to the placement of the lateral system in the structure.
A force coupling effect will be produced when asymmetrical structures are loaded
laterally. The effect cannot be captured in a 2D model.

1.3 Summary
The main objective of this research is to derive an optimization scheme to help
designers to generate economical steel frame structures in the conceptual design
phase. This optimization scheme reduces the effort designers need for preliminary
design. The "multi-level design" approach adopted from Liang's paper (2000) is re-
created in this paper with some modifications. The approach uses two optimization
loops in which the first loop generates the steel frame beam and column structure
under strength constraints. The second optimization loop generates optimal lateral
bracing systems in steel frame structures using a derived bracing removal criterion.

13
The removal criterion removes lateral bracing elements that are underutilized. The
final outcomes of the optimization scheme are considered more economical as
compared to designs made using conventional methods. This is proven using a
performance index derived while studying the results of the study.

1.4 Thesis Outline

The thesis is divided into five chapters. Chapter Two covers the literature review of
optimization techniques used in structural engineering. Chapter Three records the
methodology used in the research in two sections. The first part describes two
computational tools used in the research, MATLAB and SAP2000. The functions of
each of the computational tools used in the study are explained. The second part of
Chapter Three explains in detail the optimization scheme involved in the study,
including the first and second optimization loop.

In Chapter Four, a performance index is developed and two case studies are
conducted. In both case studies, the performance index is used as a comparison tool
between the designs derived from the optimization scheme proposed in the study and
the designs made using conventional methods. Finally, Chapter Five presents the
findings of the study and proposes possible future work.

14
2.0 Chapter 2 - Literature Review
2.1 Outline
This chapter reviews the origins of optimization in structural engineering. The type of

optimization done in the field of structural engineering is explored here, and the
process of developing the optimization scheme in the study is discussed.

2.2 Optimization Techniques in Structural Engineering

The use of numerical optimization techniques to design in the field of structural

engineering has been recorded since 1956. One of the seminal studies of optimization
techniques for frame structures was done by Heyman in 1956 (Smith 2002).

Optimization techniques shorten the time consumed by designers in creating better

conceptual structures. Conventional methods of designing lateral bracing systems are
based on trial-and-error methods with the aid of previous experience (Liang 2000).
This process of designing requires a deep understanding of the nature of structures, in
which case intuition from experience can sometimes be incorrect. Humans can make
mistakes, and designs developed using the conventional method must be tested
thoroughly to ensure safety. Much of the mentioned tedious processes involved in the
conventional method of design can be mitigated using an optimization scheme.

In the field of structural engineering, optimization techniques have been used to solve
many different problems, especially in design. These problems cover every aspect of
structural engineering, ranging from the smallest structural elements such as a bolt
up to the whole structure. These optimization techniques can be generally grouped
into three distinct categories, which are cross-sectional optimization, topology
optimization and geometry optimization (Smith 2002). The cross-sectional optimization
focuses on sizing structural elements by assuming fixed topology and geometry. The
sizing of the structural elements in this manner is approached using methods such as
performance-based design or strength-based design (Connor 2003). The topology
optimization studies the placement of structural elements in a design. One of the
common techniques used in topology optimization is the element removal technique
based on stress limits (Smith 2002). Geometry optimization is a technique that
combines the usage of both the cross-sectional optimization and the topology

15
optimization. The "multi-level design" technique used in the study is an optimization
scheme that is adopted from Liang's previous research (2000), which is categorized as
geometry optimization.

Many building optimization techniques have been developed in the past years by
various researches. In the past, the usage of optimization techniques in the field of
structural engineering is done using alternative techniques. They change based on the
objective of optimization desired by designers. In the research paper by Adeli and
Karim (1997), a neural network model is used to optimize the shapes and sizes of cold-
formed steel against its respective compliances. The neural dynamic model developed
by Adeli and Park (used the Lyapunov function to help prove stability of the
optimization process and finds the local minimum of the structural optimization
problem. In another paper, Baker (1992) used energy methods to measure the
efficiency of structural elements and Zhou and Rozvany (1992) improved the efficiency
of optimization schemes on sizing large scale structural systems using discretized
optimality criteria (Liang 2000).

All the mentioned techniques of optimization are used in the cross-sectional

optimization category in the field of structural engineering. Those techniques only
work if the topology of the structural system is fixed. In all the design problems, the
placements of structural elements in the building design are to be determined by the
designers, especially when designing the lateral system of structures. The optimization
of the topology of a lateral bracing system is difficult due to the highly non-linear
nature of structures and the large number of possibilities for the arrangement of
bracing elements (Connor 2003). Pure rigid frames are not economical to resist lateral
loading compared to diagonal bracing truss elements (McCormac 2011), but the
placement of the diagonal members are crucial to make the lateral resisting system
economical.

Many papers write about different methods of topology optimization on 2D structural

lateral bracing schemes to achieve the same final result, for example, a paper by Liang
(2011), a paper by Baldock (2006) and a paper by Kicinger (2004). A result obtained
from these papers is illustrated in Figure 1. These papers mostly utilize performance

16
base optimization that achieves the minimum compliance in a structure under lateral
loading using the minimum amount of materials. This is because the design of a
multistory steel frame structure is usually controlled by performance rather than
strength (Connor 2003). This proves the consistency of the different optimization
techniques used in those papers. Some of these papers use only 2D models while
others further develop their optimization technique to accommodate 3D space. 2D
models are only able to idealize symmetrical buildings with symmetrical loads, 3D
models are able to capture more realistic structural responses, such as force coupling
effects from unsymmetrical loadings (Connor 2003).

Figure 1: Optimized topology layout of a 2D steel frame structure. (Liang 2000)

In the study, only linear static wind load is investigated. The dynamic effects from
wind and seismic are neglected. Dynamic loading is not being investigated in this
study mainly because dynamic wind loading does not control design of short
structures. This can be seen in paper by Mijar (1998), where the steel frame structures
are optimized according to static compliance minimization and eigenvalue optimization.
The eigenvalue optimization includes the effect of dynamic loading in the structural
optimization, and the study yields the same result as Liang's paper (2000). These
assumptions allow the computational time of the optimization process to be much
faster.

The "multi-level design" optimization scheme in this study has been adopted from
Liang's paper (2000). In Liang's paper, the optimization scheme involves two loops for
steel frame structures in a linear time path. The first optimization loop designs the
gravity system of steel frame structures based on strength constraints. The sized steel

17
frame structure from the first optimization loop is used as the starting point in the
second optimization loop. The second optimization loop generates the optimal topology
of a lateral bracing system for multistory steel frame structures by using a frame
removal criterion. The frame removal criterion gradually removes inefficient materials
with the lowest strain energy from the continuum design domain. A similar frame
removal criterion is used in another paper by Querin (1999). Querin uses a more
robust bi-directional evolutionary structural optimization which combines frame
removal in addition to optimizing the topology of bracing systems in steel frame
structures.

All papers mentioned previously state that topology optimization techniques mainly
focus on deriving optimized structural bracing schemes that satisfy the building
design code limits. A more adaptable optimization scheme is proposed in this paper. A
space constraint is introduced in the second loop of the optimization scheme. The
similar type of constraint can be found in a paper by Smith (2002). Smith proposed
two different types of obstacle constraints, which are one-sided planar and spherical
constraints. The one-sided planar constraint is used to keep all structural elements on
one side of a planar surface defined. The spherical constraint is used to define a
spherical null space with structural elements. The application of both space
constraints is shown in Figure 2. These space constraints allow the designer to specify
null spaces in the structure. In our study, no bracing elements will be placed in the
null spaces. These null spaces allow for large areas in the structure to provide for
large windows for natural lighting.

Figure 2: Left: Tripod designed using limited by the sphere volume constraint.
Right: A bridge created limiting all trusses beneath the deck. (Smith 2002)

18
In conclusion, many optimization research papers have been written for design of steel
structures. Each of these papers has its own interpretation and implementation
scheme. The variations are in terms of the optimization schemes, types of materials,

types of loadings, methods of evaluating compliance, and constraints. In this study,

precedent studies are incorporated into the investigation process and a logical
optimization scheme is adopted. The major contributionof this study is the addition of a
space constraintthat differentiates this researchfrom previous precedents.

19
20
3.1 Chapter 3 - Methodology
3.2 Outline
This chapter discusses the computational tools involved in the study. It begins by
describing the overall flow of processes involved in optimizing the conceptual models of
conventional steel frame structures for more economical design solutions. The process
consists of two optimization loops, which have been adopted from a previous research
paper (Liang et al., 2000). The first optimization loop generates optimal multistory
unbraced frame structures. The second optimization loop gives economical solutions
of structural bracing system topologies. A spatial constraint is embedded into the
second optimization loop and will be further described in this chapter.

3.3 Computational Tool

The entirety of this study is conducted using two computational tools, MATLAB and
SAP2000, to automate processes of iteration. MATLAB stores and runs the algorithms
for all the optimization processes involved (First Optimization Loop and Second
Optimization Loop). SAP2000 generates and analyzes the structural models using the
inputs from MATLAB. An intercommunication platform between the two computation
tools is developed using the documented SAP2000's Application Programming
Interface (API). The API can be found in the installation file of SAP2000. The details of
each function for both computational tools used in the study are described in the
following section.

3.3.1 Optimization Algorithms - MATLAB

In the study, the first function of MATLAB is to store and execute codes to convert
parametric design variables into arrays of numerical data. These arrays of numerical
data are used to define, generate and modify the structural models in SAP2000. The
second function of MATLAB is to store and execute user-defined algorithms to iterate
the optimization variables whenever needed. These algorithms were then used to
remove under-utilized bracing elements and generate adequate initial guesses for
optimization. The third function of MATLAB used in the study is the inbuilt
optimization toolbox for both the first optimization loop and the second optimization

21
loop. The MATLAB codes constructed for this study were recorded in the Appendix.
The codes consist of two optimization loops with two inbuilt function each.

Within the optimization toolbox, the "fmincon" algorithm has been utilized as an
optimization algorithm that finds the minimum outcome of a constrained nonlinear
multivariable objective function. The 'fmincon" algorithm in the optimization toolbox
has a limitation that only accepts objective functions that produce one scalar quantity
each. In the study, the objectives of both of the optimization loops were to minimize
the total weight of the structure. While using the 'fmincon" algorithm, equality or
inequality constraints, which can be linear or nonlinear, have to be defined to set the
limits for the optimization solutions. For the same reason, upper bounds and lower
bounds of the solutions can also be specified for the 'fmincon" algorithm. Although the
structural analysis used is linear, the geometry of structures can be nonlinear (Connor
2003), the strength constraints in the first optimization loop and the compliance
constraints in the second optimization loop are both nonlinear inequalities.

One criterion while using the "fmincon" algorithm is that designers are required to use
engineering judgment to guess for the initial inputs of the optimization variables. The
'fmincon" algorithm tends to converge to a solution that is the local minimum of a
function. By having a "good" initial guess of the optimization, in which case the
variables are close to the optimal solution, the convergence rate will be much higher.
However, a "good" initial guess is not straightforward, and skills for appropriate
guesses are developed through many years of engineering experience.

3.3.2 Structural Analysis Software - SAP2000

In the study, the main function of SAP2000 computational structural analysis tool is
to carry out structural analysis on the steel frame structure using static frame
analysis. The simplest structural analysis approach has been chosen to shorten the
total amount of time required to optimize the topology of the bracing scheme. The
main structural response that is needed for the optimization are the total weight of the
structure, the compliance of the structure under defined load cases and the force and
moment diagrams in each member of the structure. Once all the optimization
processes were completed, SAP2000 is used to generate graphic structural model.

22
3.4 Optimization Scheme
This study investigates an optimization scheme to generate economical and realistic
structural bracing schemes under static lateral loadings. The optimization scheme
consists of two optimization loops, aiming to quickly create novel and physically
realistic steel frame structures. The first optimization loop is created to generate
optimal beam and column sizes in a steel frame structure. In the second optimization
loop, braces will be generated by joining all the points that made up the vertical planes
and an algorithm will be used to gradually remove the underutilized braces.

Optimization Slchemre

Problem Specification

First Optimization Loop

Second Optimization Loop

Figure 3: Flowchart of general processes involved in the optimization

scheme (detailed process of first optimization loop and second optimization
loop are shown in Figures 3 and 5).

The general processes involved in the optimization scheme are shown in Figure 2.
More detailed flowcharts for the first optimization loop and the second optimization
loop will be shown in Figure 4 and Figure 6. To use the optimization scheme,
designers are required to input all the ten design parameters to specify the problem.
These design variables have no bounds, but the designers have input physically
realistic parameters to obtain a feasible structure. These inputs should also be scaled
to the units specified in the scheme accordingly. The design parameters and their
respective units used by the scheme are as follows:

23
 1) Number of stories.
2) Number of frontal bay.
3) Number of side bay.
4) Height of one of the floor. [ft]
5) Width of one of the frontal bay. [ft]
6) Width of one of the side bay. [ft]
7) Number of floors that are grouped together to use a same steel member size for
beams and columns
8) External loading - Superimposed Dead Load, Live Load and Wind Load
magnitude. [k/ft]
9) Thickness of concrete slabs. [ft]
10) Space constraints parameters.

The first six parameters define the overall geometry of the multistory steel frame
structure. The seventh parameter is an option provided to generate more realistic
results. This is because in a typical structural design process, designers group
structural elements that experience similar stress to save time and cost for both
design and construction (McCormac, 2011). During the design phase, sizing every
structural element is very time consuming and not economical. Alternatively, the
typical method to design is to group similar elements together and size the controlling
element. In the construction phase, if elements are sized with great variation, a lot of
mistakes will occur while pieces of elements are being moved around on the
construction site. The seventh parameter also needs to be within a reasonable range,
for example, refrain from having group of ten floors. Within each group, the beams and
columns are further broken down into smaller groups, interior/exterior beams and
interior/exterior columns. The exterior elements take on twice as much load as their
interior counterpart.

The eighth parameter defines the external loading applied on the structure. The live
load magnitude depends on the purpose of the structure. The superimposed dead load
is mainly created by the concrete slabs. The wind load will be a function of location
and height of the structure. The ninth parameter is the thickness of the concrete slab
throughout the whole structure. The thickness of the concrete slab mainly controls the

24
lateral stiffness provided by the slab to the structure. The last user input parameter is
the space constraint parameter. The space constraint parameter is defined as
coordinates that bound a space that will be free of bracings.

3.5 First Optimization Loop

Once all the parameters are input into MATLAB, the first optimization loop begins. A
complete process flowchart of the first optimization loop is recorded in Figure 4. The
first optimization loop starts out as MATLAB sets up the intercommunication platform
with SAP2000 using the documented API codes. MATLAB passes data to SAP2000 as a
one-dimensional array. Once the platform is set up, MATLAB will prompt SAP2000
application to begin and generate a blank new space for modeling. The default units
used in SAP2000 are inches for length and kips-force for force. Similarly, the scale of
magnitude of forces defined in civil structural problems is in kips-force, but the unit
used to define these problems is usually in feet. For the ease for unity of units used to
define data points, the default units in SAP2000 are reset to kips-force for force and
feet for length. In this study, the only material used in the study is ASTM A572-50
structural steel, with minimum yield strength of 50ksi. A square cross section of 6
inches by 6 inches is defined as the default frame section. The choice of a square cross
section is to simplify calculations in the study that will be discussed later.

Once the intercommunication platform between MATLAB and SAP2000 is built and
the modeling space is defined accordingly, the constructed MATLAB sequence will
automatically generate points and frame elements using the Cartesian coordinate
system. The first six parameters input by the users are used to generate arrays of data
that define the location of the points and frame elements on a Cartesian coordinate
system. The points generated in SAP2000 help to define the location where each frame
element intersects and are also used to reference points whenever needed. To reiterate,
the frame elements are the beams and columns in the steel frame structure. Once the
frame elements are generated, another set of algorithms will group these frame
elements into smaller groups according to the nature of loads they are sustaining. The
whole structure will be stratified into layers of groups of floors according to the
seventh parameter input by designers. In each of these stratified layers, the frame
elements are further broken-down into four groups, which are the interior or exterior

25
beams and the interior or exterior columns. The grouping of frame elements will be
helpful when applying load or section modifiers to the frame elements.

Initialing SAP2000 using MATLAB )

Defining Units1 Materials Frame & Section Properties

Generate Points, Frame Elements using Cartesian Coordinate System

Group Frame Elements According to their Locations

(Release Moment Fixity of all Beams and Assign Base Restraints

Define and Apply Loadings

E
."" Objective- minimize weight
Constraint- strength criteria
Variable- section modifers

(Assign a Section Modifiers to Each Group ofSteel Beams and Columns

Analyze the Model

CheckConstraint

Figure 4: The first optimization loop process flowchart.

In the study, the beams in the steel frame structure are all modeled as simply-
supported members. To model a simply-supported beam in SAP2000, moment is
released from both ends of the beam and torsional fixity is released only at one end of
the beam for stability. With this configuration, the beam becomes less effective in the
lateral support system. Using the points defined, every point located on the ground
level is restrained to idealize the pinned support condition. The final step before using
the "fmincon"algorithm on the structure is applying loads to the structure. The load

26
magnitudes are parameters that are provided by the designers. The live load should be
computed according to the functionality of the building. The superimposed dead loads
are mostly contributed from the concrete slab. The self-weight of all the steel members
in the structure are computed by SAP2000. As for the wind load, a linear wind profile
is assumed, varying from zero at the base of the structure to the magnitude specified
by designers at the top of the structure.

Once all the geometry, loads and material properties have been defined into SAP2000,
the parameters in 'fmincon" algorithm are to be specified next. In the "fmincon"
algorithm, the objective is set to minimize the weight of the structure. The constraint
is set to ensure that each frame element satisfies the axial force and moment strength
criteria, and the variable to be optimized is the section modifier.

Frme Proper tfns Modi~catOn Factors

Property/Stifness Modiirs for Analysis -

Cros-section (amiaq Area

Shear Area in2 diection
Shear Areain 3 diecdion
Torsional Constant
Moment of Inertia abotA 2 ai*
Moment of Inertia abotA 3 axis
Man

Figure 5: Section modifier user interface in SAP2000 required eight variables input.

The section modifier is chosen to be the variable in this optimization algorithm for two
reasons. First, it is much easier to apply a section modifier on structural elements in
SAP2000 as compared to the process of redefining new sections. Section modifiers are
easily defined in SAP2000 modeling space; while the process of redefining new
sections requires them to be defined and old sections to be replaced. When new
sections are defined, the old sections do not get deleted but instead pile up in the

27
sections bank when the optimization process iterates. The second reason that the
section modifier is chosen is because they are able to modify all the frame element
section properties at once. For the first optimization loop, the weight of structures
(objective of optimization) is only influenced by the cross sectional area of the steel
members used since the material and geometry of the beams and columns are fixed
attributes defined by the designers. The satisfaction of strength criteria (constraint of
optimization) in each steel member is directly influenced by the area and moment of
inertia of a section. All these factors that influence the objective and the constraint of
the "finincon"algorithm are directly correlated to the section modifier.

In this study, section modifiers are consistent within groups of steel members to
simulate realistic design. The user interface for section modifiers in SAP2000 required
eight variable inputs as shown in Figure 5. For simplification of computation, the
study uses a square cross sectional area with the assumption that the depth and
width of the steel section vary at the same rate. The section modifiers influence each
member directly as follow:

Xgroup = section modifier of a group (1)

If all the sections are square,

base = width = L (2)

rate of change in base = rate of change in width = Xgroup (3)

Then the section modifiers for each group of elements are as follow,

Cross section (axial)areagroup = (Xgroup) 2 (4)

Shear area in 2 directiongroup = (Xgroup) 2 (5)
Shear area in 3 directiongroup = (Xgroup) 2 (6)
Torsional constantgroup= (Xgroup) 4 (7)
Moment of inertia about 2 axisgroup = (Xgroup) 4 (8)
Moment of inertiaabout 3 axisgroup = (Xgroup) 4 (9)
Massgroup = (Xgroup)2 (10)

28
 2 (11)
Weightgroup = (Xgroup)

The following parameters to be defined in the 'fmincon" algorithm are the initial
guesses for the variables, which give the lower and upper bound of the variables. The
initial guess for the variable of the "fmincon"algorithm is set up according expected
load path in the steel frame structure, frame elements supporting larger tributary area
are sized larger. The beams and columns are grouped together based on their
respective heights in the structure, a few floors are grouped together and the sizes
decrease towards the top of the structure. Within this height determined group of
frame elements, members are further distinguished in size based on their proximity to
the exterior of the structure. This is because beams or columns that are located at the
exterior bound of the structure only support half the tributary area compared to their
interior counterparts. The lower and upper bound of the variables are not necessary
within this study because the variables are bounded by the strength criteria constraint.

Once all the "fmincon" algorithm parameters are defined, the algorithm begins to
execute. First, "fnincon"algorithm will apply the initial guess of section modifiers to
their respective groups of frame elements and perform static frame analysis on the
structure. The objective of the "fmincon"algorithm, which is the weight of the frame
skeletal structure, is computed by SAP2000 by considering only unfactored self-weight
of every frame element. The program stores the value of the weight computed and
checks the constraint on each of the frame elements to ensure that they are all within
their elastic strength region. For all the beams in the steel frame structure, the
maximum moments are computed using SAP2000 and compared to the moment
capacity in each beam. The moment capacity for each beam is calculated using the
following equation (AISC 2011):

Moment capacity = - Fy - Z, (12)

Z = widthxdepth
2 (13)
x6

where, 0 = safety factor in steel moment capacity, 0.90

Fy = steel elastic yielding strength, 50ksi

Z,, = plastic section modulus for square cross section

29
As for the columns, the axial stress in each column is computed using SAP2000 and
compared to the axial load capacity. Since all the columns are in compression, the
steel strength of each column is determined as the minimum value between Euler's
buckling strength and the steel elastic yielding strength (AISC 2011). The axial force
load capacity in each column is calculated using the following formula:

2
F = minimum ,(F) (14)

Axial load capacity = 0 A, - F (15)

where, E = elastic modulus of steel, 29000ksi
I = moment of inertia
k = column effective length factor
L = length of column

0 = safety factor for axial load capacity in steel, 0.90

As = cross sectional area of steel

Using the MATLAB inbuilt 'fmincon" algorithm, the initial guess of section modifiers
are required to yield a result that satisfies the constraint. Once the first calculation is
completed, the "fmincon" algorithm uses the Active-Set algorithm to iterate the next set
of variables. The Active-Set algorithm has been chosen because the gradient used in
linear static frame analysis is not provided in the objective function. The Active-Set
algorithm computes the next iteration of variable using a "quasi-Newton
approximation to the Hessian of the Lagrangian" (The Math Works, Inc. 2013). This
allows the Active-Set algorithm to take large steps between iterations, allowing for
faster convergence of results. The Active-Set algorithm computes the Lagrange
multipliers of the objective function and stores them. All subsets of Lagrange
multipliers with infeasible constraints are removed. The more detailed process of the
Active-Set algorithm can be found in the book, "Numerical Optimization", by Jorge
Nocedal (2006) and the paper, "A fast algorithm for nonlinearly constrained
optimization calculations" by Michael Powell. Once the minimum objective of the first
optimization loop has been achieved, the second optimization loop begins.

30
3.6 Second Optimization Loop
The output from the first optimization loop, which is the set of the section modifiers,
applied on all the beams and columns for the gravity system of the steel frame. Similar
to the first optimization loop, a flow chart of the processes involved in the second
optimization loop are developed and illustrated in Figure 6.

Unlock SAP2000 Model

Add in Cross-Braced Steel Truss Element To Represent Concrete Slab

Add in cross-braced Steel Truss Element to Represent Concrete Slab

Add in Bracing Elements in All Vertical Plane

Group Bracings

Remove Bracings According to the Space Constraints

Set up Initial Variables for Function

E Define:
Objective- minimum weight
Constraint- com liant

Set Modifer

ERun Model

Get Force and Moment in Bracing Elements

Remove Bracing with Usage s 0.01%

Bracing Removed?)

YES NO

Figure 6: The second optimization loop process flowchart.

31
The second optimization loop is developed to generate economical topologies of the
building bracing scheme for the lateral support system. While generating the
economical topologies, some assumptions are made to simplify the structural design
process. In this study, only linear static wind loading is considered. All the steel
bracing elements are considered to be truss elements, which do not resist any
shearing or twisting action. Before placing all the bracings in the steel frame structure,
the SAP2000 model has to be unlocked; this is a safety feature in SAP2000 to ensure
that the structural model is unaffected when the analysis is running. The first
addition to the structural steel frame in the second optimization loop is to place cross-
braced steel frames into the structure to substitute for the stiffness provided by
reinforced concrete slabs. The method used to model these cross-braced steel trusses
is recorded in the following section.

3.6.1 Equivalent Frame Analysis for Reinforced Concrete Slab

The goal to maintain high physical accuracy in the structural modeling encounters a
problem when concrete slabs are to be included in the structural model for lateral
loading resistance. Concrete slabs in a multistory steel frame structure contribute a
significant portion of lateral stiffness to the structure and cannot be ignored while
modeling a realistic structure (Corley 1961). Typically in a computational structural
modeling space, concrete slabs are modeled as thin shells or plates with the equivalent
stiffness (Asif 2013). The structural response of thin shells or plates cannot be
computed using the static frame analysis. A more sophisticated finite element analysis
is required to be used and the analysis will consume considerably more time. In order
to keep the structural analysis process fast and simple, the concrete slabs cannot be
modeled as thin shells or plates.

A method of idealizing concrete slabs as simple steel frame elements was proposed in
the paper, "The Equivalent Frame Analysis for Reinforced Concrete Slabs", by Corley
in 1961. The method measures the lateral stiffness provided by concrete slabs in a
multistory steel frame structure and replaces all the concrete slabs using cross-braced
steel truss elements. The method is illustrated in Figure 7. The method is adopted in
our study.

32
 Figure 7: The equivalent frame transformation that convert the stiffness in a
concrete slab to two cross-braced steel truss elements.

First, the torsional stiffness of the concrete slabs that contributes to the lateral
stiffness is calculated using the follow formulas (AISC 318-05):

Ktorsional - 9Ec 3(16)

L 2 (1- C2IL2)

C = - .63x X3Y (17)

where: Ec = Young's modulus of elasticity for concrete

C = torsional constant
L2 = width of the concrete slab

C2 = width of columns supporting the concrete slab

x = shorter side of the rectangular concrete cross section

y = longer side of the rectangular concrete cross section

In the study, the concrete slabs are assumed to be normal-weight concrete with a

specified 28-day compressive strength of 4ksi. The Young's modulus of elasticity of the
concrete in psi was calculated using the following formula:

Ec = 57,000V (18)

where: fc' = specified 28-day compressive strength of concrete [psi]

In the denominator portion of the calculation for the torsional stiffness, the equation
can be further simplified by assuming that the optimization scheme always minimizes
the width of the columns supporting the concrete slab, but limited by the strength

33
constraints. Therefore, c2 will have considerably lower value than L2 . Hence, the ratio
of C2 /L can be neglected in the Ktorsional equation.

The end result from equation 16 is the lateral stiffness provided by the concrete beam
at a direction that is parallel to the direction of L2. The stiffness, Ktorsional , does not
translate directly as the stiffness of the cross-braced truss elements because of the
difference in direction. A geometry transformation equation can be applied to Ktorsional
to get the required stiffness in each cross-braced truss element, which is directly
converted to the required steel cross sectional area. The area required in each cross-
braced truss element can be derived as follow (Connor 2003):

V = 2- F- cos 6 (19)
V = Ktorsional - Ugiobai (20)

F= Atrussrequired-Esteel (21)
Ltruss

Uglobal -COSO = uloca (22)

Ktorsionai Ltruss
Atruss_required 2 2
= -Atrussrequired-EsteeCos g(

where: V = applied force

F = axial force in the truss element
6 = angle between the truss element and the direction of L 2

Uglobal = global displacement

Atrussrequired = required area in each truss element

Esteel = Young's modulus of elasticity for steel (29,000ksi)

Ltruss = length of truss element

Using this entire equivalent frame modeling method, the computational time can be
greatly reduced.

3.6.2Second Optimization Loop (continued)

Once all the cross-braced steel truss elements are placed into the steel frame
structure, the steel frame structure will be flooded by braces. All these bracing
elements are first placed between two points that are not in the line of any beam or

34
column. These pairs of points are ones that have at least two different coordinates in
terms of x, y or z. Then, all the joints that are different in all x, y and z coordinates are
removed from the system as the optimization algorithm is meant to generate realistic
steel frames where no braces should run across open room spaces. At the same time,
all braces that are in the horizontal plane are also removed because there should not
be braces running across the slab system in a real steel frame structure. The last
modification to the bracing elements is to remove all the bracings according to space
constraints. The space constraint is defined in a vector form with six elements as
follow:

bottom f loor number

top f loor number
space constraint - frontal wall ending number (24)

side wall startingnumber

side wall ending number

These six elements bound a space that is free of bracings elements. There is no
limitation to the number of space constraints that can be defined in the algorithm, but
having too many null spaces with no bracings might make the structure unstable. The
whole process of defining and gradual removal of all the bracing elements in the steel
frame structure is illustrated in Figure 8. The process of defining all the steel braces is
embedded in the second optimization loop. Once all the steel braces are in place, all
the braces in the structure will be redefined as truss elements by releasing their
moment fixity. After the removal of the bracing elements due to impracticability and
space constraints, there still exist many bracing elements leftover in the frame
structure. If one variable is assigned to each leftover bracing element, the optimization
algorithm will take a very long time to be executed. Therefore, those leftover bracing
elements are grouped together according to their location, length and nature of force
applied on them. The bracing elements are located higher in the structure are smaller
than those that are nearer to the base. The longer the bracing elements, the larger its
cross sectional area. Since steel frame structures are not always symmetrical from the
front and the side view, the bracing elements should be sized accordingly.

35
 Figure 8: Process of
gradually removing
bracing elements.
First, bracing removal
to obtain realistic
bracing placement
and then to satisfy
space constraint.

36
Once the structural elements that are required for the lateral system analysis are
modeled in SAP2000, the second "fmincon"algorithm in MATLAB will be set up. In the
second optimization loop, the objective of the "fmincon"algorithm is similar to the

objective in the first "fmincon"algorithm, which is to minimize the total weight of the
whole structure. Since the weight of the structure is directly related to the amount of
material required to build the structure, the solutions from the second "fmincon"
algorithm can be considered economical in terms of material usage. As for the
constraint of the 'fmincon" algorithm, compliance in the lateral movement of the
structure is investigated. The variables of the second "fnincon"algorithm are set to be
section modifiers, similar to the first 'fmincon" algorithm, but the section modifiers in
the second "fmincon" algorithm are applied on the groups of bracings instead of all the
beams and columns. Once these three parameters of the second "finircon"algorithm
are defined, the initial guesses, upper bound and lower bound of the variable will be
inputted. The initial guesses for the second 'fmincon" algorithm is made by the
designer while taking into consideration the mentioned sizing criterion while grouping
the bracing elements. The same rule on the upper and lower bound applied on the
first 'fmincon" algorithm is adopted in the second 'fmincon" algorithm. The bounds are
not critical as the variables are actually limited by the constraint.

The second "fmincon"algorithm starts to execute once all the required input is defined.
Initially, the second "fmincon"algorithm applies the initial guess for section modifiers
on the respective groups of bracings. Then, SAP2000 will be used to run static frame
analysis on the frame structure. The overall weight of the structure will be computed
by using the load case that only considers the self-weight of every structural element.
The lateral compliance of the structure is extracted from the result computed by
SAP2000. The constraint of the second "fmincon"algorithm is set up to compare the

inter-story drift of every joint in the structure to an inter-story drift limit of h (AISC

2011). The variable h in the equation refers to the story height. Similar to the first
"fmincon"algorithm, the result from the initial guesses in the second 'fmincon"
algorithm are required to satisfy the constraint. If the constraint is not satisfied, the
initial guesses are required to be rescaled. The same optimization algorithm, Active-Set
algorithm is used in the second 'fmincon" algorithm. Using the initial result, the

37
second 'fmincon" algorithm generates the next set of variables using the same
approach explained in the first "fmincon"algorithm.

Once the second 'fmincon" algorithm satisfies the objective of optimization while being
constrained, the resulting frame structure will be checked against a frame removing
criteria. The frame removing criteria used in this study compares the axial forces in
each bracing element to their respective axial load capacity. If the usage of a bracing
element is less than 0.01% of its axial load capacity, the algorithm will remove the
bracing from the structure. While using this frame removing criteria, the designers
have to be especially careful while grouping bracing elements. This is because only one
steel size will be assigned to each group of bracing elements based on the bracing
element that requires the largest steel size in each group. If bracing elements within
each group experiences large disparities in member forces, then those bracing
elements with low member forces will be removed unintentionally due to inefficient
material usage. This limitation of the frame removal criteria is hard to avoid but can
be minimized by further breaking down the grouping of bracing elements. The most
optimal situation is that no bracing elements are grouped together, but the time taken
for the optimization process will be inefficiently long. Since there is no way to ensure
that the frame removal criteria are perfect, the final solution of the study can only be
considered "economical" instead of "most optimized" solution.

Once all the underutilized bracings are removed from the structure, a constructed
code will regenerate new variables for a rerun of the 'fmincon" algorithm. This code
makes sure that the new variables satisfy the constraint of the 'fmincon" algorithm by
gradually increasing the magnitude of the variables. The new generated variables will
be used to initiate the second "fmincon"algorithm on the modified structure. After
some of the bracing elements in the steel frame structure are removed, the
computation time of the optimization is shown to improve drastically. This whole
process goes on until the frame removing criteria no longer removes any bracing
elements.

In the second optimization loop, some modifications were done to the 'fmincon"
algorithm to speed up the rate of convergence. The termination tolerance of the

38
objective and the compliance were adjusted to be 0.01, instead of the default value of
le-6. This is acceptable because an error of 5% is considered negligible in structural
design. Safety factors placed in pertinent design codes have already considered and
precluded these negligible errors (ASIC 2011).

39
40
4.0 Chapter 4- Results
4.1 Outline
Four case studies have been investigated using the developed optimization scheme. A

performance index is developed to compare each result from all four cases to steel
frame designed using a conventional method.

4.2 Performance Index

The ultimate objective of this study is to generate economical conceptual design of
steel frame structures. The measurement of the objective of the study can be done by

comparing costs from the results generated using the optimization scheme or a
conventional method. The total cost of a structure ties together with many factors,
including the complexity of design (design time), efficiency of design (amount of
material) and constructability of design (construction time). A performance index is
developed to measure the value of a design considering factors mentioned above.

The complexity of design is a subjective issue that varies in everyone's opinion. The
comparison of the complexity of design is not included in the performance index
However, the optimization scheme developed in this study only requires the designer
to decide on the parametric design variables. These parametric design variables are
usually provided by the building owner. The rest of the steps involved in steel frame
structure design are all tackled in the proposed optimization scheme.
Sizing the beams and columns is a tedious process but it is done in the first
optimization loop in the scheme. Designers usually use trial-and-error methods with
their engineering experience to find a solution for the placement of the lateral bracing
element in a steel frame structure. The process can take up a long time and the result
can be feasible but with room for improvement. The second optimization loop is able to
derive economical solutions of steel frame structures in a much shorter time.

In terms of efficient of design, total cost of a structure is determined by its type and
amount of materials used. In this study, the material investigated is ASTM572-50
3
structure steel with a mass of 4901b/ft (McCormac 2011). Since the type of material
is fixed, the performance index developed takes into account of only the amount of

41
material used. The amount of material used is determined from the total weight of the
structure.

The same gravity system will be utilized in the design by the optimization scheme and
the design through conventional methods. The constructability of the structure is
measured by numbers of lateral bracing elements. As the number of bracing elements
increase, the effort required to build the structure increases. The effort can be
quantified by the difficulty to place each bracing elements, but difficulty is a subject
matter that is hard to be used to compare the designs. In the performance index, the
constructability of the steel frame structure is quantified by the number of
connections required to be built. Connections are costly and have a direct influence on
the final cost of a design (McCormac 2011). Each bracing element requires two
connections to be installed on into the steel frame structure. The total number of
connections is calculated based on doubling the amount of bracing elements.

The two factors used to derive the performance index measure different quantities.
Equivalent cost values of each element measured in the two factors is calculated by
using the 2013 market value. For the cost of the material, a ton of steel costs about
700 US$ (Steelonthenet.com 2013). This is equivalent to about 320 US$ for a kip of
steel. For the connections, the cost of a connection is a function of the force provided
by the connection. For simplicity of calculation, all the connections are assumed to be
the same and have a cost of 50 US$/connection. The result of the performance index
will calculated as follow:

PerformanceIndex = (total weight[kips] x 320 -+(# connections x 50 [conection]) (25)

4.3 Results
In this thesis, two case studies were examined. In each case study, a similar steel
frame structure is made using the conventional design methods. Since the
optimization scheme takes into account the space constraints, frames designed using
the conventional method are also limited by the same constraints. The space
constraints specified 2D null spaces at all the exterior walls of the structure. The
space constraints should be consistent when doing the comparison because the study

42
examines the optimal placement of the bracing elements and placing bracing elements
furthest away from the center of rigidity will be the most optimal placement of bracing
elements (Connor 2003). Therefore, no bracing elements should be placed in the null

spaces in using both design method. While designing the steel frame structure using
the conventional method, the placement of the bracing elements is determined using
trial-and-error. Steel frame structures designed using the conventional method initiate
steel frame models derived from the first optimization loop. Instead of going through
the second optimization loop, all the bays in the steel frame structure that are not
under the space constraints are filled with cross-braced steel trusses. The compliance
of the steel frame structure is computed using SAP2000 and compared to the limited
compliance from building design code. Then the trusses in bracing the steel frame
structure are removed one by one until the compliance of the structure no longer fulfill
the design code limit. The frame removal starts from the center of the structure as
those trusses provide the least amount of the lateral stiffness to the structure.

The final steel frame structures designed using the conventional method are compared
to the steel frame derived from the optimization scheme. The weight of steel frame
structures, number of the bracing elements and compliance of both steel frame
structures are recorded for each case study. The performance index for all the steel
frame structures are computed and compared to their counterpart in each case study
graphically.

Case Study I
In the first case study, the steel frame structure is two stories tall with three frontal

bays and two side bays. All the bay widths are set to be 10 feet apart and the story
height is also 10 feet tall. Since the steel frame structure has only two stories, the
gravity system in the first floor has the same members sizing as the second floor. 4
inches thick of concrete slab is used for the flooring in for the whole structure. The
loading used in case study I is tabulated in Table 1. Using the thickness of the
concrete slab, equivalent cross-braced steel trusses are placed as the floor in the
structure. These cross-braced elements are hidden in the final display picture of the
steel frame structures to allow for a clearer view of the lateral bracing elements. The
final steel frame structures produced using the two different methods are shown in

43
Figure 10. The black frame elements are the beams and columns and the red frame
elements are the bracing elements. The visual comparison of the two methods clearly
shows that the optimization scheme produces a steel frame structure that uses less
bracing elements. The sizes of each structural element group are recorded in Table 2
and the results from case study I are tabulated in Table 3. Table 3 shows a slight
increase in weight from the conventional method to the optimized solution. The
number of connections used in the design made using conventional method is twice
the number of the connections used in the optimized solution. The cost index derived
previously is also tabulated in Table 3.

Loadings Magnitude [kips/ft]

Interior beams 2
Live Load
Exterior beams 1

Dead Load Program calculated

Interior beams 2.5

Superimposed Dead Load

varies from 1 kips/ft at the top of the

Wind WindLoadLinearly
Load
structure to 0 at the base of the structure

Table 1: Loading used in case study I

Equivalent cross-sectional area

Structural Elements Section Modifiers
[in 2 1

Interior columns 0.772 2.316

Exterior columns 0.458 1.374

Interior beams 0.732 2.196

Exterior beams 0.716 2.148

Diagonal members 0.158 0.474

Table 2: Sizes of structural elements after optimization

Weight [kips] Number of connections

Optimization 22.71 28
Scheme
Conventional 24.27 56
Method
Cost Index 320 [US$/kips] 50 [US$/ connection]
Table 3: Results of case study I

44
In Figure 9, the cost of each element in the performance index is calculated in terms of
US$. The total cost of the whole structure is computed and shown in the figure. The
steel frame structure designed using the conventional method is more costly in both
criteria examined in the performance index.

Performance Index
N Economical Solution 0 Conventional Design
12000

10000 --- -- -

8000 - - - - -_-- -- - - - -- - -

6000
0
4000 - --- - - -- - -

2000 ---

0
Materials Cost Connections Cost Total Cost

Figure 9: Performance Index comparison in Case Study I.

Figure 10: Steel frame structure of two different methods in Case Study I.

4.3.1Case Study II
In the second case study, the steel frame structure is four stories tall with three
frontal bays and three side bays. All the bay widths are set to be 10 feet apart and the
story height is also 10 feet tall. The four story structure is divided into two distinct
groups for beams and columns sizing. The first and second floors have same frame

45
member sizes and the same for the third and fourth floors. The loading used in case
study II is tabulated in Table 4. Similar to the first case study, 4 inches thick of
concrete slab is used for the flooring in for the whole structure and the equivalent
cross-braced steel trusses are placed as the floor in the structure. These cross-braced
elements are also hidden in all the final display pictures of the steel frame structures
to allow for a clearer view of the lateral bracing elements. The final steel frame
structures produced using the two different methods are shown in Figure 12. The
black frame elements are the beams and columns and the red frame elements are the
bracing elements. The denser red elements in the steel frame designed using the
conventional method shows more bracings elements placed in the structure. The sizes
of each structural element group are recorded in Table 5 and the results from case
study I are tabulated in Table 6. Table 6 shows a slight increase in weight and number
of connections used from the conventional method to the optimized solution. The cost
index derived previously is also tabulated in Table 6. In Figure 11, the cost of each
element in the performance index is calculated in terms of US$. The total cost of the
whole structure is computed and shown in the figure. The steel frame structure
designed using the conventional method is more costly in both criteria examined in
the performance index.

Loadings Magnitude [kips/ft)

Interior beams 2
Live Load
Exterior beams 1
Dead Load Program calculated
Interior beams 2.5
Superimposed Dead Load

Linearly varies from 1 kips/ft at the top of the

Wind Load
structure to 0 at the base of the structure
Table 4: Loading used in case study II.

46
 Section Modifiers Equivalent cross-sectional area
Structural Elements
[in 2 j
Interior columns
0.772 2.316
(Floor 1 to 2)
Exterior columns
(Floor 1 to 2) 0.458 1.374

Interior beams
0.732 2.196
(Floor 1 to 2)
Exterior beams
0.716 2.148
(Floor 1 to 2)
Interior columns
0.778 2.334
(Floor 3 to 4)
Exterior columns
0.65 1.950
(Floor 3 to 4)
Interior beams
0.744 2.232
(Floor 3 to 4)
Exterior beams
0.698 2.094
(Floor 3 to 4)
Diagonal members 1 bay
0.0528 0.158
(Floor 1 to 2)
Diagonal members 2 bay
0.0754 0.226
(Floor 1 to 2)
Diagonal members 3 bay
0.1006 0.302
(Floor 1 to 2)
Diagonal members 1 bay
0.1176 0.353
(Floor 3 to 4)
Diagonal members 2 bay
0.0502 0.151
(Floor 3 to 4)
Diagonal members 3 bay
0.09 0.270
(Floor 3 to 4)

Table 5: Sizes of structural elements after optimization.

47
 Weight [kip~s] Number of connections
Optimization 60.22 50
Scheme
Conventional 64.41 64
Method I I
Cost Index 320 [US$/kips1 50 [US$/connectio
Table 6: Results of case study II.

Performance Index
m Optimized Solution m Conventional Design

30000 --- -- ----

25000 ----- _ - - - -

20000 -- - -

15000 - --
0

10000 - -

5000 -- - - --

0-
Materials Cost Connections Cost Total Cost

Figure 11: Performance Index comparison in Case Study II

Figure 12: Steel frame structure of two different methods in Case Study II

48
5.0 Chapter 5 - Discussion
5.1 Outline
This chapter analyzes the results from the case studies. The future application and
the adaptability of the optimization scheme proposed in the study are explored.

5.2 Discussion
Both case studies show a similar trend of results. The steel frame structure designed
using the optimization scheme developed in the study produces more economical
structures. The designs made from the optimization scheme in both case studies use
less material in terms of weight and number of connections, and they are easier to
construct as well as being less expensive. Also, interpreting the design step of the steel
frame structure, the optimization scheme can be considered more robust because
work conducted by the designers is considerably cut. The designers are only required
to input the parametric design variables into the developed MATLAB optimization
scheme and the rest of the job will be done by the computer. As for the conventional
method, the designers are required to spend time and use previous engineering
experience to run different structural models using computational structural analysis
tools to find a "good" design. This process takes time and experience, which makes the
optimization scheme a more economical method to derive conceptual designs of the
steel frame structure.

Both designs in all the two case studies are checked to satisfy the compliance limit
given in the AISC (2011) design code. The design derived using different methods
yields different results. In a realistic structural design case, they all satisfy the code
limit and are feasible structures. However, the design derived using the conventional
method deflects less. This is because the conventional method uses more structural
bracing elements. However, in designing civil structures, all the additional material
used is wasted as the design codes have already been made to ensure safety. The
comparison can be further studied by increasing the deflection limit on the steel frame
structure designed using the optimization scheme. This further investigation of the
topic will help to determine the reliability of the optimization scheme.

49
The main objective of the research is to derive an optimization scheme to help
designers to generate economical steel frame structures in the conceptual design
phase. Using the performance index developed in this study, the results generated
using the optimization scheme are more economical than steel frame structures
designed using the conventional method. Also, in both case studies, the space
constraints successfully restrain the structures from have bracing elements on all the
external walls. This concludes that the optimization scheme derived in this study
helps to generate economical results with open spaces.

5.3 Future Work

There are some limitations to the optimization scheme proposed in the study. Many
assumptions are made in the study and if any of the assumptions are not compatible
with the real situation, the design is considered unfeasible. For example, the study is
only done on linear static lateral loadings, if the structure is located in a seismically
active area, the optimization scheme does not yield realistic results. However,
modifications can be made to the optimization scheme to cover a wider range of steel
frame engineering. The current optimization scheme is concluded to be able to
produce economical results in a small scale structure. However, if the number of
variables in the study increases, the computation time will increase exponentially and
the optimization tool becomes infeasible.

There are several modifications that can be done on the proposed optimization scheme.
For example, the frame removal criteria used in this study can only generate
economical lateral bracing schemes compared to conventional designs. The frame
removal criteria can be improved to ensure that optimal bracing schemes can be found.
A good method is to use the bi-directional evolutionary structural optimization that
allows the frame removal criteria to remove and add bracing elements. Another option
to explore would be one that makes the optimization scheme more robust by finding
ways to reduce the time consumed in optimization. With these additions,
improvements can be made in reducing the structural response computation time,
reducing the amount of variables or using a better optimization algorithm.

50
6.0 References
- Adeli, Hojjat, and Asim Karim. "Neural network model for optimization of cold-
formed steel beams." Journal of Structural Engineering 123, no. 11 (1997):
1535-1543.
- Adeli, Hojjat, and Hyo Seon Park. "A neural dynamics model for structural
optimization-theory." Computers & structures 57, no. 3 (1995): 383-390.
- Asif, Abell, Mike, "Technical Knowledge Base." CSi Knowledge Base. Last
modified Jan 16 , 2013. Accessed May 7, 2013.
https://wiki.csiberkeleyhttp://www.chicagomanualofstyle.org/tools-citationgui
de.html.com/x/u5AS
- Baker, William F. "Energy-based design of lateral systems." Structural
Engineering International 2.2 (1992): 99-102.
- Baldock, Robert, and Kristina Shea. "Structural topology optimization of braced
steel frameworks using genetic programming." In Intelligent Computing in
Engineering and Architecture, pp. 54-61. Springer Berlin Heidelberg, 2006.
- Bertsekas, Dimitri P., and John N. Tsitsiklis. "Neuro-dynamic programming: An
overview." In Decision and Control, 1995., Proceedings of the 34th IEEE
Conference on, vol. 1, pp. 560-564. IEEE, 1995.
- Connor, Jerome, Introduction to Structural Motion Control, (Pearson Education,
Inc., 2003)
- Corley, William Gene, M. A. Sozen, and C. P. Siess. "The equivalent frame
analysis for reinforced concrete slabs." PhD diss., University of Illinois at
Urbana-Champaign, 1961.
- Gaynor, Andrew T., James K. Guest, and Cristopher D. Moen. "Reinforced
concrete force visualization and design using bilinear truss-continuum topology
optimization." Journal of Structural Engineering (2012).
- Hoenderkamp, Johan CD, Hubertus H. Snijder, and Herm Hofmeyer. "Push-pull
interface connections in steel frames with precast concrete infill panels." Open
Constr. Build. Technol. J 6 (2012): 63-73.
- Kicinger, Rafal, Tomasz Arciszewski, and Kenneth De Jong. "Morphogenesis
and structural design: cellular automata representations of steel structures in
tall buildings." In Evolutionary Computation, 2004. CEC2004. Congress on, vol.
1, pp. 411-418. IEEE, 2004.
- Kim, C. K., H. S. Kim, J. S. Hwang, and S. M. Hong. "Stiffness-based optimal
design of tall steel frameworks subject to lateral loading." Structural
optimization 15, no. 3-4 (1998): 180-186.
- Lee, S., S. Bobby, S. M. J. Spence, A. Tovar, and A. Kareem. "Shape and
Topology Sculpting of Tall Buildings under Aerodynamic Loads." In 20th
Analysis and Computation Specialty Conference, pp. 323-334. ASCE, 2012.
- Liang, Qing Quan, and Grant P. Steven. "A performance-based optimization
method for topology design of continuum structures with mean compliance

51
 constraints." Computer methods in applied mechanics and Engineering 191, no.
13 (2002): 1471-1489.
- Liang, Qing Quan, Yi Min Xie, and Grant P. Steven. "Optimal topology design of
bracing systems for multistory steel frames." Journal of Structural Engineering
126, no. 7 (2000): 823-829.
- Manual, Steel Construction. "American Institute of Steel Construction." AISC,
USA (2011).
- McCormac, Jack C., Stephen F. Csernak, and Manojkumar V. Chitawadagi.
Structural steel design. Intext Educational Publishers, 2011.
- Mijar, Anand R., Colby C. Swan, Jasbir S. Arora, and Iku Kosaka. "Continuum
topology optimization for concept design of frame bracing systems." Journal of
Structural Engineering 124, no. 5 (1998): 541-550.
- Murty, Katta G. Linear complementarity, linear and nonlinear programming.
Berlin: Heldermann, 1988.
- Nocedal, Jorge, and Stephen J. Wright. Numerical optimization. Springer
Science+ Business Media, 2006.
- Powell, Michael JD. "A fast algorithm for nonlinearly constrained optimization
calculations." In Numerical analysis, pp. 144-157. Springer Berlin Heidelberg,
1978.
- Querin, 0. M., G. P. Steven, and Y. M. Xie. "Evolutionary structural
optimization (ESO) using a bidirectional algorithm." Engineering Computations
15, no. 8 (1998): 1031-1048.
- Querin, 0. M., V. Young, G. P. Steven, and Y. M. Xie. "Computational efficiency
and validation of bi-directional evolutionary structural optimisation." Computer
methods in applied mechanics and engineering 189, no. 2 (2000): 559-573.
- Rozvany, G. I. N., M. Zhou, and T. Birker. "Generalized shape optimization
without homogenization." Structural Optimization 4, no. 3-4 (1992): 250-252.
- Steelonthenet.com, "Steel prices for years 2012 and 2013." Last modified 2013.
Accessed May 12, 2013. http://www.steelonthenet.com/steel-prices.html.
- Smith, Jeffrey, Jessica Hodgins, Irving Oppenheim, and Andrew Witkin.
"Creating models of truss structures with optimization." In ACM Transactions
on Graphics (TOG), vol. 21, no. 3, pp. 295-301. ACM, 2002.
- Stromberg, Lauren L., Alessandro Beghini, William F. Baker, and Glaucio H.
Paulino. "Application of layout and topology optimization using pattern
gradation for the conceptual design of buildings." Structural and
Multidisciplinary Optimization 43, no. 2 (2011): 165-180.
- Tenek, Lazarus H., and Ichiro Hagiwara. "Static and vibrational shape and
topology optimization using homogenization and mathematical programming."
Computer methods in applied mechanics and engineering 109, no. 1 (1993):
143-154.
- The Math Works, Inc., "Documentation of fmincon function (Active-Set-
Algorithm)." Last modified 2013. Accessed May 9, 2013.
http://www.mathworks.com/help/optim/ug/fmincon.html.

52
Appendix

MATLAB Codes

53
5/22/13 3:45 AM C:\Users\John\Deskt... \Optimization5.m 1 of 14

%% Clean up the workspace of MATLAB

clear;
clc;
format long;
%%Initiate the SAP2000 API, constructing the platform that links SAP2000 with MATLABe
as described in the chapter 3.3 (Computation Tool)
for nl = 1;
% pass data to Sap2000 as one-dimensional arrays
feature('COMSafeArraySingleDim', 1);
% pass non-scalar arrays to Sap2000 API by reference
feature('COMPassSafeArrayByRef', 1);
% create Sap2000 object, any ERROR, please exit from SAP2000 fully
SapObject = actxserver('sap2000.SapObject');
% create SapModel object
SapModel = SapObject.SapModel;
% start Sap2000 application
SapObject.ApplicationStart;
% initialize model
ret = SapModel.InitializeNewModel;
% create new blank model
ret = SapModel.File.NewBlank;
end
%% Set the units used in SAP2000 as kips for force and ft for length
for n1 = 1;
ret = SapModel.SetPresentUnits(4);
end
%%Input parametric design (Specifying problem) - described the First Optimization Loops
Chapter under the Methodology
numberofstory = 7;
numberoffrontal bay = 4;
numberofside bay = 4;
height_of_story = 10; %ft
widthoffrontal bay = 10; %ft
width of side bay = 10; %ft
group of story = 3; % number of story which the sizes of steel sections arek/
grouped together
%assume that cross section is a square
widthofcrosssection = 0.5; %[ft]
A = (widthofcrosssection)^2; %cross sectional area of initial section [ft^2]
thickness_of_concreteslab = 4/12; %4 inches concrete slab[ft]
f c = 4; %[ksi]
E c = 57*sqrt(fc*1000); %[ksi]
SUPERDEAD = 1; %[kip/ft]
LIVELOAD = 1; %[kip/ft]
WINDLOAD = 5; %[kip/ft]
number_of_groupof floor = fix((numberofstory-)/groupofstory)+1;
totalbeamsandcolumns = ((numberofstory)*(numberoffrontalbay+1)*l
(number of side bay+1))+((number of story)*(number of frontal bay)*I/
(number of side bay+l))+...
((numberofstory)*(numberofsidebay)*Il
(numberoffrontal bay+l)); %total number of beams and columns
% space constraints input variables;
% input format of wall with no bracings = [bottomfloor# topfloor# FVwallstart#L/
FVwallend# SVwallstart# SVwallend#];
% 1 <= bottomfloor# <= topfloor# <= number of story
54

..
.............
.....
....
.....
.........
. ....
..
..
.. ...........
5/22/13 3:45 AM C:\Users\John\Deskt... \Optimization5.m 2 of 14

% 1 <= FVwallstart# <= FVwallend# <= (numberoffrontal bay + 1)

% 1 <= SVwallstart# <= SVwallend# <= (numberofsidebay + 1)
% input should not remove all the bracings in any floors
removal(1,:) = [1 7 1 5 1 1);
removal(2,:) = [1 7 1 5 5 5];
removal(3,:) = [1 7 1 1 1 51;
removal(4,:) = [1 7 5 5 1 5];
% current setting take out all exterior bracings

%% Defining materials of structure as Steel

for n1 = 1;
ret = SapModel.PropMaterial.AddQuick('A572Gr5O',1,5,0,0,0,0,0,'A572Gr5O');
end
%%DEFINE square frame section property
for nl = 1;
ret = SapModel.PropFrame.SetRectangle('.5ftx.5ft rectangle','A572Gr5O', L'
width ofcrosssection,widthofcrosssection);
end
%% ADD point object by coordinates generated from the parametric design variables byk'
designers
for n1 = 1;
for n=1:((numberoffrontalbay+1)*(numberofsidebay+1)*(numberofstory+));
ret = SapModel.PointObj.AddCartesian((mod(n-l,numberoffrontal-bay+l))*/
(width offrontalbay), ...
mod((fix((n-1)./(number offrontal bay+1))),number-of-side-bay+1)*'
(width of side bay), ...
(fix((n-1)./((numberoffrontalbay+l)*(numberof side bay+1))))*1'
(heightofstory),'','');
end
totalpoints = ((numberoffrontalbay+l)*(numberofsidebay+l)*hL
(numberofstory+l));
total_pointsin one floor = ((numberof frontal bay+l)*(numberofsidebay+1));
end
%% ADD frame object by coordinates generated from the parametric design variables byV'
designers(beams and columns)
for n1 = 1;
for n=l:((numberof story)*(numberoffrontalbay+l)*(numberof side bay+l));
ret = SapModel.FrameObj.AddByCoord((mod(n-l,numberoffrontalbay+l))*L/
(widthoffrontalbay), ...
mod((fix((n-1)./(number of frontal bay+l))),number-of-side-bay+l)*L/
(width of side bay), ...
(fix((n-1)./((numberoffrontalbay+l)*(numberof side bay+l))))*W'
(heightofstory), ...
(mod(n-1,numberoffrontalbay+l))*(widthoffrontalbay),
mod((fix((n-1)./(numberoffrontal bay+l))),numberofside bay+l)*1'
(widthofsidebay), ...
(fix((n-1)./((numberof frontalbay+l)*(numberofsidebay+l)))+l)*V,
(height of story), ...
', 'lftxlft rectangle', sprintf('Column %d',n), 'GLOBAL'); %vertical frameV
elememts
end

for n=1:((number of story)*(numberoffrontal bay)*(numberofside bay+l));

[ret, a] = SapModel.FrameObj.AddByCoord((mod(n-,number offrontalbay))*l
(width of frontalbay),
55
 5/22/13 3:45 AM C:\Users\John\Deskt.. .\0ptimization5.m 3 of 14

mod((fix((n-1)./(numberoffrontalbay))),numberofsidebay+l)*'
(widthofsidebay), ...
(fix((n-1)./((numberoffrontal bay)*(numberofsidebay+l)))+1)*le
(height_of_story), ...
(mod(n-1,numberoffrontalbay)+1)*(width of frontalbay),
(mod((fix((n-1)./(numberoffrontalbay))),numberofside bay+i))* '
(widthofside bay), ...
(fix((n-1)./((numberoffrontalbay)*(numberofsidebay+i)))+i)*L'
(heightofstory), ...
', 'lftxlftrectangle', sprintf('FV Beam %d',n), 'GLOBAL'); %front viewW'
horizontal frame elememts
end

for n=1:( (number-of-story)*(number-of-side-bay)*(numberof frontal bay+1));

[ret, a) = SapModel.FrameObj.AddByCoord ((mod((fix((n-1)./I'
(numberofside bay))),numberoffrontal bay+l))*(widthoffrontalbay),
((mod(n-1,numberof side bay)))*(widthofside-bay), ...
(fix((n-1) ./((numberofside bay)*(numberoffrontalbay+1)))+1)*k'
(heightofstory), ...
(mod((fix((n-1)./(numberofside bay))),numberoffrontal bay+1))* '
(widthoffrontal bay), ...
((mod(n-1,numberof side bay))+1)*(widthofside bay),
(fix((n-1)./((numberofsidebay)*(numberoffrontalbay+1)))+1)*L'
(heightofstory), ...
'', 'lftxlftrectangle', sprintf('SV Beam %d',n), 'GLOBAL'); %side viewll
horizontal frame elememts
end
end
%%ADD and assign group for all beams and columns to allow the optimization scheme tok/
generate relistic structure - described in page 29 of thesis
for ni = 1;
ret = SapModel.GroupDef.SetGroup('Interior Column');
ret = SapModel.GroupDef.SetGroup('Exterior Column');
ret = SapModel.GroupDef.SetGroup('Exterior Column Front');
ret = SapModel.GroupDef.SetGroup('Exterior Column Back');
ret = SapModel.GroupDef.SetGroup('Exterior Column Left');
ret = SapModel.GroupDef.SetGroup('Exterior Column Right');
ret = SapModel.GroupDef.SetGroup('Interior Beam');
ret = SapModel.GroupDef.SetGroup('Exterior Beam');

for n=i:((numberof_story)*(numberoffrontal bay+1)*(numberofside bay+i));

ret = SapModel.FrameObj.SetGroupAssign(sprintf('Column %d',n),'Interior 1
Column',false(),0);
end

for m=1:numberofstory;
for n=l:numberoffrontalbay+1;
ret = SapModel.FrameObj .SetGroupAssign (sprintf ('Column %d',(n+(m-i)*L'
(numberoffrontal bay+1)*(numberofside bay+1))),'Exterior Column Front' ,false(),0);
ret = SapModel.FrameObj.SetGroupAssign(sprintf('Column %d', (n+(m-1)*'
(numberoffrontal bay+1)*(numberofside bay+i))),'Interior Column',true(),0);
end
end

56 for m=1:numberofstory;

.............. ...............
.......
.
5/22/13 3:45 AM C:\Users\John\Deskt...\Optimization5.m 4 of 14

for n=((numberoffrontal bay+1)*(numberofsidebay)+l):I'

((numberof_frontal bay+1)*(numberofsidebay+l));
ret = SapModel.FrameObj.SetGroupAssign(sprintf('Column %d',n+(m-i)*K/
(number of frontal bay+1)*(number of side bay+l)),'Exterior Column Back',falseo,O);
ret = SapModel.FrameObj.SetGroupAssign(sprintf('Column %d',n+(m-i)* '
(numberof_frontal bay+l)*(numberofsidebay+l)),'Interior Column',true),0);
end
end

for m=l:number of story;

for n=l:number of frontal bay+1:((number of side bay+1)*L/
(number of frontal bay+l));
ret = SapModel.FrameObj.SetGroupAssign(sprintf('Column %d',n+(m-i)*L/
(numberoffrontal_bay+l)*(numberofside_bay+i)),'Exterior Column Left',false(,O);
ret = SapModel.FrameObj.SetGroupAssign(sprintf('Column %d',n+(m-1)*L'
(numberoffrontal bay+l)*(number of side bay+1)),'Interior Column',true(),O);
end
end

for m=l:number of story;

for n=(numberoffrontal bay+l):numberoffrontalbay+l:L/
((numberofside_bay+l)*(numberoffrontalbay+));
ret = SapModel.FrameObj.SetGroupAssign(sprintf('Column %d',n+(m-1)*L(
(numberoffrontal bay+)*(numberofsidebay+l)),'Exterior Column Right',false(,0);
ret = SapModel.FrameObj.SetGroupAssign(sprintf('Column %d',n+(m-i)*L'
(numberoffrontal bay+1)*(number ofsidebay+i)),'Interior Column',trueo,O);
end
end

ret = SapModel.FrameObj.SetGroupAssign('Exterior Column Right','Exterior V

Column', false() ,1);
ret = SapModel.FrameObj.SetGroupAssign('Exterior Column Left','Exteriork'
Column',false(),1);
ret = SapModel.FrameObj.SetGroupAssign('Exterior Column Front','ExteriorkM
Column',false(),l);
ret = SapModel.FrameObj.SetGroupAssign('Exterior Column Back','Exteriork'
Column',false(),1);

for m=1:numberofstory;
for n=1:numberoffrontalbay;
ret = SapModel.FrameObj.SetGroupAssign(sprintf('FV Beam %d',n+(m-i) 1'
*numberof_frontal bay*(numberofsidebay+1)),'Exterior Beam',false),0);
end
for n=l+number of frontalbay*numberofside bay:number of frontalbay*L/
(number of sidebay+l);
ret = SapModel.FrameObj.SetGroupAssign(sprintf('FV Beam %d',n+(m-i) L'
*numberoffrontalbay*(numberofsidebay+1)),'Exterior Beam',falseo,O);
end
for n=l:numberofsidebay;
ret = SapModel.FrameObj.SetGroupAssign(sprintf('SV Beam %d',n+(m-1)*il
(number of frontalbay+l)*numberof side bay),'Exterior Beam',falseo,O);
end
for n=1+numberoffrontalbay*numberofsidebay:(numberoffrontal bay+1) L'
*number of side bay;
ret = SapModel.FrameObj.SetGroupAssign(sprintf('SV Beam %d',n+(m-1)*u'
57
5/22/13 3:45 AM C:\Users\John\Deskt... \Optimization5.m 5 of 14

(numberoffrontalbay+l)*numberofside bay),'Exterior Beam',falseo,O);

end
end

for n=l:numberoffrontal bay*(numberof side bay+l)*numberofstory;

ret = SapModel.FrameObj.SetGroupAssign(sprintf('FV Beam %d',n),'Interior
Beam',false() , 0);
end

for n=1:(numberoffrontal_bay+l)*numberofside bay*numberofstory;

ret = SapModel.FrameObj.SetGroupAssign(sprintf('SV Beam %d',n),'Interiork/
Beam',false(),0);
end
ret = SapModel.FrameObj.SetGroupAssign ('Exterior Beam', 'Interior Beam', L/
true(),1);
end
%% Continue grouping of beams and columns
for n1 = 1;
total number of columns in one floor=(number of frontal bay+l)*(number of side-bay+l);
totalnumberofFVbeamsinonefloor=(numberoffrontal bay)*'
(numberofsidebay+l);
totalnumberofSVbeamsinonefloor=(numberoffrontal bay+l)*L/
(numberofside bay);
for h=l:group_of_story:numberofstory;
if h<numberofstory-groupof story+1;
ret = SapModel.GroupDef.SetGroup(sprintf('Floor %d to %d Interior Column',h,L
h+groupofstory-1));
ret = SapModel.GroupDef.SetGroup(sprintf('Floor %d to %d Exterior Column',h,
h+groupofstory-1));
ret = SapModel.GroupDef.SetGroup(sprintf('Floor %d to %d Interior Beam',h,L
h+groupofstory-1));
ret = SapModel.GroupDef.SetGroup(sprintf('Floor %d to %d Exterior Beam',h,L(
h+group of story-1));
for n=(h-l)*totalnumberofcolumnsinonefloor+1:(h-l+groupofstory) V
*totalnumberofcolumnsinonefloor;
ret = SapModel.FrameObj .SetGroupAssign (sprintf ('Column %d',n),sprintf ('Floor %dL
to %d Interior Column',h,h+group of story-1),false(),0);
ret = SapModel.FrameObj.SetGroupAssign(sprintf('Column %d',n),sprintf('Floor %dI
to %d Exterior Column',h,h+groupofstory-1),false(),0);
end
ret = SapModel.FrameObj.SetGroupAssign('Exterior Column',sprintf('Floor %d to %V
d Interior Column',h,h+group of story-1),true(),l);
ret = SapModel.FrameObj.SetGroupAssign('Interior Column',sprintf('Floor %d to %
d Exterior Column',h,h+groupofstory-1),true(),1);
for n=(h-1)*totalnumberofFVbeamsinonefloor+1:(h-+groupofstory)
*totalnumberofFVbeamsinonefloor;
ret = SapModel.FrameObj.SetGroupAssign(sprintf('FV Beam %d',n),sprintf ('Floor %V
d to %d Interior Beam',h,h+groupofstory-1),false(),0);
ret = SapModel.FrameObj.SetGroupAssign(sprintf('FV Beam %d',n),sprintf ('Floor %L/
d to %d Exterior Beam',h,h+groupof_story-1),falseo,0);
end
for n=(h-l)*totalnumberof_SVbeamsinonefloor+l:(h-+groupofstory)L/
*total number ofSV beams in one floor;
ret = SapModel.FrameObj .SetGroupAssign (sprintf ('SV Beam %d',n),sprintf ('Floor %V
d to %d Interior Beam',h,h+group_of_story-1),falseo,0);
58
5/22/13 3:45 AM C:\Users\John\Deskt...\Optimization5.m 6 of 14

ret = SapModel.FrameObj.SetGroupAssign(sprintf('SV Beam %d',n),sprintf('Floor %V

d to %d Exterior Beam',h,h+group-ofstory-1),falseo,0);
end
ret = SapModel.FrameObj.SetGroupAssign('Exterior Beam',sprintf('Floor %d to %d
Interior Beam',h,h+group_of_story-1),true() ,1);
ret = SapModel.FrameObj.SetGroupAssign('Interior Beam',sprintf('Floor %d to %d
Exterior Beam',h,h+group_of_story-1),true() ,);
elseif h>=numberofstory-groupof_story+1;
ret = SapModel.GroupDef.SetGroup(sprintf('Floor %d to roof Interior Column',V
h));
ret = SapModel.GroupDef.SetGroup(sprintf('Floor %d to roof Exterior Column',W/
h));
ret = SapModel.GroupDef.SetGroup(sprintf('Floor %d to roof Interior Beam',h));
ret = SapModel.GroupDef.SetGroup(sprintf('Floor %d to roof Exterior Beam',h));
for n=(h-1)*totalnumberofcolumnsinonefloor+l:L
number ofstory*totalnumber ofcolumnsinonefloor;
ret = SapModel.FrameObj.SetGroupAssign(sprintf('Column %d',n),sprintf('Floor %dv
to roof Interior Column',h),falseo,O);
ret = SapModel.FrameObj.SetGroupAssign(sprintf('Column %d',n),sprintf('Floor %dL1
to roof Exterior Column',h),false(),O);
end
ret = SapModel.FrameObj.SetGroupAssign('Exterior Column',sprintf('Floor %d toV1
roof Interior Column',h),true(),l);
1
ret = SapModel.FrameObj.SetGroupAssign('Interior Column',sprintf('Floor %d to'
roof Exterior Column',h),true(),1);
for n=(h-l)*totalnumberofFVbeamsinonefloor+1:(h-l+group_of_story)'1
*total number ofFV beams in one floor;
ret = SapModel.FrameObj.SetGroupAssign(sprintf('FV Beam %d',n),sprintf('Floor %/
d to roof Interior Beam',h),falseo,O);
ret = SapModel.FrameObj.SetGroupAssign(sprintf('FV Beam %d',n),sprintf('Floor %V
d to roof Exterior Beam',h),false(),O);
end
for n=(h-l)*total number ofSVbeams in one floor+l:V
numberofstory*totalnumberofSVbeamsinonefloor;
ret = SapModel.FrameObj.SetGroupAssign(sprintf('SV Beam %d',n),sprintf('Floor %L/
d to roof Interior Beam',h),falseo,O);
ret = SapModel.FrameObj.SetGroupAssign(sprintf('SV Beam %d',n),sprintf('Floor %/
d to roof Exterior Beam',h),falseo,O);
end
ret = SapModel.FrameObj.SetGroupAssign('Exterior Beam',sprintf('Floor %d tok/
roof Interior Beam',h),trueo,l);
ret = SapModel.FrameObj.SetGroupAssign('Interior Beam',sprintf('Floor %d toLl
roof Exterior Beam',h),true(),l);
end
end
end
%% SET release at all beams
for nl = 1;
null = (zeros(6,1,'double'));
strongaxismoment = logical([0;0;0;0;0;1]);
ret = SapModel.FrameObj.SetReleases('Interior Beam', strongaxismoment,l
strongaxismoment, null,null,1);
ret = SapModel.FrameObj.SetReleases('Exterior Beam', strongaxismoment,L/
strongaxismoment, null,null,l);
end
59
5/22/13 3:45 AM C:\Users\John\Deskt . .\Optimization5.m 7 of 14

%% ASSIGN point object restraint at base

for nl = 1;
fixedbase = logical(ones(6,1));
pinned = logical([1;1;1;0;0;0]);
roller = logical([0;0;1;0;0;0]);
11
ret = SapModel.PointObj.DeleteRestraint('ALL',1); % delete all existing
restraints
for n=l:((number of frontal bay+1)*(number of side bay+l));
ret = SapModel.PointObj.SetRestraint(num2str(n),pinned);
end
end
%% ADD load patterns
for nl = 1;
ret = SapModel.LoadPatterns.Add('SUPERDEAD',2,0, true());
ret = SapModel.LoadPatterns.Add('LIVE',3,0, true();
ret = SapModel.LoadPatterns.Add('WIND',6,0, true());
ret = SapModel.LoadPatterns.AutoWind.SetASCE705('WIND',1, 0, 0.8, 0.5, 5, 0.15, V
0.15, false(, 0, 0, 70, 1, 1, 1.1, 0.85, 0.85);
end
%% ASSIGN loading for load pattern
for nl = 1;
ret = SapModel.FrameObj.SetLoadDistributed('Interior Beam' , 'SUPERDEAD', 1, 10,V
0, 1, SUPERDEAD, SUPERDEAD,'GLOBAL',true(),false(),l);
ret = SapModel.FrameObj.SetLoadDistributed('Interior Beam' , 'LIVE', 1, 10, 0,L/
1, LIVELOAD, LIVELOAD,'GLOBAL',true(),false(),1);
ret = SapModel.FrameObj.SetLoadDistributed('Exterior Beam' , 'SUPERDEAD', 1, 10,W/
0, 1, SUPERDEAD, SUPERDEAD,'GLOBAL',true(),false(),l);
ret = SapModel.FrameObj.SetLoadDistributed('Exterior Beam' 'LIVE', 1, 10, O,11
1, LIVELOAD, LIVELOAD,'GLOBAL',true(),false(),l);
for m=l:number of story;
for n=l:numberoffrontalbay;
ret = SapModel.FrameObj.SetLoadDistributed(sprintf('FV Beam %d',n+((m-l)Z
*totalnumberofFVbeamsinonefloor)), ...
1
'WIND', 1, 8, 0, 1, m/number of story*WINDLOAD,
m/number of-story*WINDLOAD,'GLOBAL',true(),false(),0);
end
for n=1:numberofside-bay;
ret = SapModel.FrameObj.SetLoadDistributed(sprintf('SV Beam %d',n+((m-1)V11
*totalnumberofSVbeamsinonefloor)), ...
'WIND', 1, 7, 0, 1, m/number of story*WINDLOAD,v
m/numberofstory*WINDLOAD,'GLOBAL',true(),false(),0);
end
end
end
%% ADD load combinations
for n1 = 1;
ret = SapModel.RespCombo.AddDesignDefaultCombos(true(), false(), false(),
false(); %add steel frame design default combo
% delete load case 1,3,4,6,7,8
% load case 2 for gravity design
% load case 5 for lateral design
ret = SapModel.RespCombo.Delete('UDSTL1');
ret = SapModel.RespCombo.Delete('UDSTL3');
ret = SapModel.RespCombo.Delete('UDSTL4');

60 ret = SapModel.RespCombo.Delete('UDSTL6');
5/22/13 3:45 AM C:\Users\John\Deskt...\Optimization5.m 8 of 14

ret = SapModel.RespCombo.Delete('UDSTL7');
ret = SapModel.RespCombo.Delete('UDSTL8');
end
%% REFRESH view
for nl = 1;
ret = SapModel.View.RefreshView(0, false());
end
%% The first "fmincon" optimization algorithm by MATLAB - described further in page 30Z
of thesis
for nl = 1;
%making xO as [n+l ; n ; n+1 ; n .......... ; 2 ; 1 ; 2; 1]
number_ofgroupof_floor = fix((numberofstory-)/groupofstory)+l;
xO=ones(numberofgroupoffloor*4,1);
for n = 2 : numberofgroupof floor;
x0(1:4*(n-1),1)=x0(1:4*(n-1),1)+ones(4*(n-1),1);
end
for n = 1 : number_of_groupof floor;
x0(1+((n-1)*4):4+((n-1)*4),1)=x0(1+((n-1)*4):4+((n-1)*4),1)+[1;0;1;0];
end
lb=zeros(numberofgroup of floor*4,1);
ub=max(x0)*ones(number_ofgroupoffloor*4,1);
options = optimoptions('fmincon');
options = optimoptions(options,'DiffMinChange', 0.5);
options = optimoptions(options,'Display', 'iter');
options = optimoptions(options,'TolCon', le-2);
options = optimoptions(options,'TolFun', le-2);
options = optimoptions(options,'ToiX', le-2);
x = fmincon(@BeamColumnSizing2,x0, [], [], [],[],b,ub,@StrengthConstraints2,.1
options)
[Weight] = BeamColumnSizing2(x);[c,ceq]=StrengthConstraints2(x);
end
%% UNLOCK model
for nl = 1;
ret = SapModel.SetModelIsLocked(false());
end
%%The beginning of the second optimization loop.
% ADD in all cross braced frame elements to represent concrete slab
for ni = 1;
% read all the coordinates of the existing points
X=zeros(1,1, 'double') ;Y=zeros(1,1, 'double') ;Z=zeros(1,1, 'double') ;L/
a='';
matrix = zeros(totalpoints,3);
for n=1:total_points;
[ret, matrix(n,l), matrix(n,2), matrix(n,3)] = SapModel.L1
PointObj.GetCoordCartesian(sprintf('%d',n), X, Y, Z);
end
% calculation equivalent frame element area
torsionalconstantfrontal = (1-0.63/
*thickness of concrete slab/width of frontal bay)*V
(thicknessofconcreteslabA3*widthoffrontal bay/3);
torsionalconstantside = (1-0.63/
*thicknessofconcreteslab/widthofsidebay)*/
(thicknessofconcreteslab^3*widthofsidebay/3);
K_t_frontal = (9*Ec*torsionalconstant frontal)/widthoffrontalbay;
K t side = (9*E c*torsional constant side)/width of side bay;
61
5/22/13 3:45 AM C:\Users\John\Deskt...\Optimization5.m 9 of 14

theta = atan(widthofsidebay/widthoffrontal bay);

lengthof crossbraces = widthofside bay/cos(theta);
areaofcrossbraces = K_t_frontal*length of crossbraces/(2*29000*(cost
(theta))^2);
area modifiers for cross braces = area of cross braces/A;
% setting modifiers
ModValue = zeros(1,1,'double');
ModValue(1,1) = (areamodifiersforcrossbraces); %cross sectionalt
area modifier
ModValue(2,1) = (areamodifiersforcrossbraces); %shear area in localz
2 direction modifier
ModValue(3,1) = (areamodifiersforcrossbraces); %shear area in localL/
3 direction modifier
ModValue(4,1) = (areamodifiersforcrossbraces)^2; %Torsional'
constant modifier
ModValue(5,1) = (areamodifiersforcrossbraces)^2; %Moment of inertiak
about local 2 axis modifier
ModValue(6,1) = (areamodifiers for cross braces)^2; %Moment of inertiak
about local 3 axis modifier
ModValue(7,1) = 0; %Mass modifier
ModValue(8,l) = 0; %Weight modifier
%set up group for cross braces
ret = SapModel.GroupDef.SetGroup('Cross braces');
% generating frame elements using coordinates
iteration = 1;
for h = 2 number of story+l;
for m = 1 numberofsidebay;
for n = 1 numberoffrontal bay;
mm = n+((m-1)*(numberoffrontal bay+l))+((h-l)*L/
(total_pointsinonefloor));
nn = n+l+((m)*(numberoffrontal bay+l))+((h-l)*/
(totalpointsinonefloor));
ret = SapModel.FrameObj.AddByCoord (matrix (mm, 1), matrix(mm, 2) ,matrix (mm,
3),...
matrix(nn,l), matrix(nn,2),matrix(nn,L6
3),
'', 'lftxlft rectangle', sprintf('CrossL'
brace %d',iteration), 'GLOBAL');
ret = SapModel.FrameObj.SetModifiers(sprintf('Cross brace %d',iteration), k6
ModValue(1:8,1),0);
ret = SapModel.FrameObj.SetGroupAssign(sprintf('Cross brace %d',
iteration),'Cross braces',false(),0);
iteration = iteration + 1;
mm = n+l+((m-l)*(numberoffrontal bay+1))+((h-l)*I/
(totalpointsinonefloor));
nn = n+((m)*(numberoffrontal bay+l))+((h-1)*(totalpointsinone_floor));
ret = SapModel.FrameObj.AddByCoord (matrix (mm, 1), matrix(mm, 2) ,matrix (mm, L
3) ,...
matrix(nn,1), matrix(nn,2),matrix(nn,V
3),
'', 'lftxlftrectangle', sprintf('Crossk/
brace %d',iteration), 'GLOBAL');
ret = SapModel.FrameObj.SetModifiers(sprintf ('Cross brace %d',iteration), e
ModValue(1:8,1),0);
ret = SapModel.FrameObj.SetGroupAssign(sprintf('Cross brace %d',L/
62
5/22/13 3:45 AM C:\Users\John\Deskt...\Optimization5.m 10 of 14

iteration),'Cross braces',false(),O);
iteration = iteration + 1;
end
end
end
ret = SapModel.FrameObj.SetReleases('Cross braces', strongaxismoment,W.
strongaxismoment, null,null,1);
total number of cross braces = iteration - 1;
totalbeamsand columnsandcrossbraces = ((numberofstory)*V/
(numberoffrontalbay+1)*(numberofsidebay+l))+((number_of_story)*/
(numberof_frontalbay)*(number of side bay+l))+...
((numberofstory)*(numberofsidebay)*l
(number of frontalbay+l))+totalnumberofcrossbraces; %total number ofk
beams and columns
end
%%beginning of the frame removal criteria (user defined algorithm)
1
% ADD in all possible bracings in the vertical planes within the floor group and group
them according to their length
for nl = 1;
% add groups for bracing according to their number of bays (equivalent to length)
for h = 1:groupof_story:number_of_story;
for n = 1:max([numberoffrontalbay,number ofsidebay,number-ofstory]);
ret = SapModel.GroupDef.SetGroup(sprintf('Bracings Group %d Above Floor %d',n,
h));
end
end
% add a frame bracing element to every two point elements that are different in atL1
least two axis of coordinate
% only braces in vertical are generated
item = 1;
for h=l:groupofstory:numberof_story;
hl = (h-l)*heightofstory;
h2 = (h+group_of_story-l)*heightofstory;
for n = 1:total_points-1;
for m = n+l:total_points;
if matrix(n,l)==matrix(m,l) && matrix(n,2)-=matrix(m,2) && matrix(n,3)-=matrixkl
(m,3) && hl<=matrix(n,3) && matrix(n,3)<=h2 && hl<=matrix(m,3) && matrix(m,3)<=h2||...
matrix(n,l)-=matrix(m,l) && matrix(n,2)==matrix(m,2) && matrix(n,3)~=matrixW/
(m,3) && hl<=matrix(n,3) && matrix(n,3)<=h2 && hl<=matrix(m,3) && matrix(m,3)<=h2 ;
ret = SapModel.FrameObj.AddByCoord(matrix(n,l), matrix(n,2),matrix(n,3),matrix'1
(m,l),matrix(m,2),matrix(m,3),...
'', 'lftxlftrectangle', sprintf('Bracings %d',item), 'GLOBAL');
if matrix(n,l)==matrix(m,l)
nn=max([(abs(matrix(n,2)-matrix(m,2))/widthoffrontal-bay), (abs(matrix(n,3)?
-matrix(m,3))/heightofstory)));
elseif matrix(n,2)==matrix(m,2)
nn=max( [ (absr(matrix(n,1)-matrix(m,1) ) /width-of-side-bay), (abs (matrix (n,3)-
matrix(m,3))/heightof story)]);
end
ret = SapModel.FrameObj.SetGroupAssign(sprintf('Bracings %d',item),sprintfLI
('Bracings Group %d Above Floor %d',nn,h),falseo,0);
item = item+l;
end
end
end
63
5/22/13 3:45 AM C:\Users\John\Deskt...\Optimization5.m 11 of 14

end
item = item-1;
for n = 1:item;
ret = SapModel.FrameObj .SetReleases ( sprintf ('Bracings %d',n), strongaxismoment,
strongaxismoment, null,null,0);
end
end
%% REMOVE bracings accordings space constraints given by designers
for nl = 1;
for n = 1:size(removal(:,l));
z1l = (removal(n,l)-l)*heightofstory;
z22 = removal(n,2)*height_ofstory;
xl = (removal(n,3)-l)*width of frontal-bay;
x22 = (removal(n,4)-l)*widthoffrontalbay;
yll = (removal(n,5)-l)*widthofsidebay;
y22 = (removal(n,6)-l)*widthofsidebay;
ret = SapModel.SelectObj.ClearSelection;
ret = SapModel.SelectObj.CoordinateRange(xl*12, x22*12,yl 1*12,y22*12,z11*12,V
z22*12,false(),'GLOBAL',false(),false(),true));
ret = SapModel.SelectObj.Group('Interior Beam',true());
ret = SapModel.SelectObj.Group('Interior Column',true());
ret = SapModel.SelectObj.Group('Exterior Beam',true());
ret = SapModel.SelectObj.Group('Exterior Column',true());
ret = SapModel.SelectObj.Group('Cross braces',true());
ret = SapModel.FrameObj.Delete('',2);
end
ret = SapModel.SelectObj.ClearSelection;
end
%% REFRESH view
for n1 = 1;
ret = SapModel.View.RefreshView(0, false());
end
%%Second "fmincon" algorithm that keeps running until no further frame removal is donev
- decribed in page 37 and 38 of thesis
for nl = 1;
NumberofFrames = 1; NumberofFramesl = 0;
while NumberofFrames -= NumberofFrames1;

% set up yO as [1,2,3,4,. ,2,3,4,5,.....3,4,5,6,.....]

number of groupof bracingsin eachfloorgroup=max([numberoffrontalbay,
number of side bay,number of story));
yO=ones V
(number of groupof floor*numberofgroupof bracingsin eachfloorgroup,1);
for n = 2 : number ofgroupof floor;
y0(1:numberof_groupofbracings_in_eachfloor_group*(n-1),1)=y0(1:K'
numberofgroupof bracingsin eachfloorgroup*(n-1),1)+onesK'
(number of groupof bracingsineachfloorgroup*(n-1),1);
end
for n = 1 : number ofgroupof floor;
for m = 1 : number of groupof bracingsin_eachfloor group;
yO((m)+(n-1)*(number_ofgroupof bracings_ineachfloorgroup),1)=y0((m)+(n-1)*1
(number of groupof bracingsin each floor_group),1)+(m-1);
end
end

64
5/22/13 3:45 AM C:\Users\John\Deskt.. \Optimization5.m 12 of 14

%making sure that constraints are always feasible

[Weight] = BracingTopologyRemoveMemberMinForce4(y0);[c,ceq) =L
ComplianceConstraints3(yO);
while max(c)>0;
yO = yO+ones(size(yO,1),1)
[Weight] = BracingTopology_RemoveMemberMinForce4(yO);[c,ceq] =L
ComplianceConstraints3(yO);
end
y0

%set upper and lower bound of y0

lb=0*ones(size(yO,1),1, 'double');
ub=2*max(y0)*ones(size(y0,1),1,'double');

% optimization
options = optimoptions('fmincon');
options = optimoptions(options,'DiffMinChange', 0.5);
options = optimoptions(options,'Display', 'iter');
options = optimoptions(options,'TolCon', le-2);
options = optimoptions(options,'TolFun', le-2);
options = optimoptions(options,'ToIX', le-2);
y = fmincon(@BracingTopologyRemoveMemberMinForce4,yO, [],[], [], [],lb,ub,
@ComplianceConstraints3,options)
% % % y = ...
% % % [0.041136296873365;...
%% % 0.054848395831153;...
% % % 0.068560494788942;...
%% % 0.082272593746730;...
%% % 7.000000000000220;...
% % % 7.999999999999996;...
%% % 9.000000000000309;...
%% % 0.033633087907873;...
%% % 0.041136296873365;...
%% % 0.054848395831153;...
%% % 0.068560494788942;...
%% % 6.000000000000229;...
%% % 7.000000000000031;...
%% % 8.000000000000025;...
%% % 0.112681369504044;...
% % % 0.027424197915577;...
% % % 0.041136296873365;...
% % % 0.073807927370027;...
%% % 5.000000000000286;...
%% % 6.000000000000012;...
%% % 7.000000000000100];
[Weight] = BracingTopologyRemoveMemberMinForce4(y);[c,ceq] =
ComplianceConstraints3(y);

% get force and moment in bracings

Name = cellstr('');
[ret,NumberofFrames,FrameNames] = SapModel.FrameObj.GetNameList(0, Name);
NumberofBracings = NumberofFrames - totalbeamsandcolumnsand cross braces;
MaxBracingAxialForce = zeros(NumberofBracings,1);

%extract current existing bracings labels

65
5/22/13 3:45 AM C:\Users\John\Deskt.. .\0ptimization5.m 13 of 14

bracingnumber = cellstr('');
for n = 1:NumberofBracings
bracing number(n,1) = regexprep(FrameNames6
(n+totalbeamsandcolumnsandcross_braces),'\D','');
end
brace = str2double(bracingnumber); %all the leftover bracing's number as/
double

%Reading the remaining bracings axial forces

ret = SapModel.Results.Setup.DeselectAllCasesAndCombosForOutput;
ret = SapModel.Results.Setup.SetComboSelectedForOutput (sprintf (I'UDSTL%d', 5));
for m = 1 : NumberofBracings;
NumberResults = 0;Obj = cellstr(' ');ObjSta = zeros(1,1,'double');Elm = cellstrW
(' ');ElmSta = zeros(1,1,'double');LoadCase = cellstr(' ');StepType = cellstr(' ');W
StepNum = zeros(1,1,'double');
P = zeros(1,1,'double');V2 = zeros(1,1,'double');V3 = zeros(1,1,'double');T =L/
zeros(1,1,'double');M2 = zeros(1,1,'double');M3 = zeros(1,1,'double');
[ret,NumberResults, Obj, ObjSta, Elm, ElmSta, LoadCase, StepType, StepNum, P,I(
V2, V3, T, M2, M3] = .
SapModel.Results.FrameForce (sprintf ('Bracings %d',brace(m)),0, NumberResults,1
Obj, ObjSta, Elm, ElmSta, LoadCase, StepType, StepNum, P, V2, V3, T, M2, M3);
MaxBracingAxialForce(m,1) = (max(P));
end

% unlock model
ret = SapModel.SetModelIsLocked(falseo);

% remove frame members if less than 0.01% capacity is in use

numberofframeremoved = 0;
for m = 1:NumberofBracings;
Modifiers = zeros(8,1,'double');
[ret,modifier] = SapModel.FrameObj.GetModifiers(sprintf('Bracings %d',bracez
(m)), Modifiers);
44
if MaxBracingAxialForce(m,1)<=0.05*0.9*50*1 *modifier(l)*A && modifier(1)<=1'
1| modifier(l)<=0.0001;
ret = SapModel.FrameObj.Delete(sprintf('Bracings %d',brace(m)),0);
numberofframeremoved = numberofframeremoved + 1;
end
end
numberofframesremoved = sprintf('%d frames',numberofframeremoved)
[ret,NumberofFramesl,FrameNames1] = SapModel.FrameObj.GetNameList(0, Name);

% refresh view
ret = SapModel.View.RefreshView(0, false());
end
[Weight] = BracingTopologyRemoveMemberMinForce4(y);[c,ceq] =Ie
ComplianceConstraints3(y);

end

66