ENG

DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO

NON-UNIFORM LOADS
by
OF TECHNOLOGY
Lavina H. Sadhwani
MAY 3 0 INSTITUTE
MASSACHUSETTS 2000
Bachelor of Science in Civil Engineering
University of Illinois at Urbana-Champaign
May 1998 LIBRARIES

SUBMITTED TO THE DEPARTMENT OF CIVIL AND ENVIRONMENTAL ENGINEERING IN

PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF

MASTER OF ENGINEERING
IN CIVIL & ENVIRONMENTAL ENGINEERING

at

MASSACHUSETTS INSTITUTE OF TECHNOLOGY

June 2000

© 2000 Lavina H. Sadhwani

All Rights Reserved.

The author hereby grants MITpermission to reproduceand distributepubliclypaper

and electronic copies of this thesis document in whole or in part.

Signature of Author I'_FW

Pepartment of Civil and Environmental Engineering

May 5, 2000

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

Accepted by
Daniele Veneziano
Chairman, Departmental Committee on Graduate Studies
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO
NON-UNIFORM LOADS
by

Lavina H. Sadhwani

Submitted to the Department of Civil and Environmental Engineering

on May 5, 2000 in Partial Fulfillment of the Requirements for the Degree of
Master of Engineering in Civil and Environmental Engineering

Abstract
Arch structures have been used for centuries in various types of structural systems, particularly buildings
and bridges. Arches are characterized by the ability to carry load primarily through axial action. An arch
shape can be optimized such that the design load pattern is carried by purely axial action. If the load pattern
on the optimized arch differs from the design loading, bending stresses develop in the arch. Load carrying
mechanisms that resist load through bending are less efficient than systems that resist load through axial
action. An active control system on an arch can be used to reduce the bending stresses in an arch. Control
forces, and resulting moments are applied to counteract moments produced various other load
combinations. The actuator, a component of the active control system, imposes counteracting control
moments on the arch to limit the resultant moment forces in the arch to a pre-defined limit.

This thesis includes a description of basic arch theory; specifically, a description of various types of arches,
force equilibrium equations for arch structures, and the theory behind design optimization of arch shapes
for a design load pattern are presented. Additionally, the fundamentals of active control systems, including
components of the system and a basic algorithm are discussed. A case study is then utilized to demonstrate
that an arch structure enabled with an active control system can carry all load patterns mainly through axial
action.

Thesis Supervisor: Jerome J. Connor

Title: Professor of Civil and Environmental Engineering
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

ACKNOWLEDGEMENTS

Writing this thesis would have taken a lot longer and been a lot more difficult if it wasn't
for a few people who provided me with guidance and support, both professionally and
personally:

ProfessorJ.J.Connor, for his guidance and advice. More importantly, for revitalizing my enthusiasm for
my career in structural engineering.

Emma Shepherdson, for helping me get started.

Frank,D.J , and Brant, my eternal appreciation for listening to my continuous nagging and whining and
for helping me with my formatting issues.

George Kokosalakis, for his assistance with the MATLAB program.

Myfamily, for their continuous support and encouragement.

Page 3 of 50
MASSACHUSETTS
XASSACHUSETTS INSTITUTE OF TECHNOLOGY
INSTITUTE OF TECHNOLOGY Page 3 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

Table of Contents

Abstract 2
1.0 Arch Structures 6
1.1 Types of Arches 7
1.2 Arch Theory 8
2.0 Funicular Method of Arch Design 11
3.0 Disadvantages of Traditional Arch Design 12
4.0 Active Control 13
4.1 Components of Active Control System 13
4.2 Actuator Technology 15
4.3 Active Control Algorithm 16
5.0 Active-Controlled Arch Structure 18
6.0 Design of Proposed Charles River Crossing Bridge Arch 19
6.1 Introduction 19
6.2 Loads 20
6.2.1 Superstructure Loads 21
6.2.2 Self-weight 23
6.3 Funicular method 24
6.3.1 Determination of Constraints 24
6.3.2 Determination of optimal arch shape 25
6.4 Computer Model 25
6.5 Optimized Arch Shape 27
6.6 Discussion of Design Stresses 28
7.0 Design of Active Control System 29
7.1 Configuration of Actuators 29
7.2 Active Control Algorithm 30
7.3 Results of Active Control Algorithm 31
8.0 Conclusion 34

TECHNOLOGY
OF TECHNOLOGY
INSTITUTE OF Page 4 ot 50
MASSACHUSETTS
MASSACHUSETTS INSTITUTE Page 4 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

Table of Figures
Figure 1: New River Gorges Bridge............................................................................................ 6
Figure 2: Natchez B ridge ......................................................................................................... 7
Figure 3: Oudry Mesly pedestrianbridge................................................................................... 8
F igure 4: Typ ical A rch .................................................................................................................... 9
Figure 5: EquilibriumForce Equationsfor an Arch................................................................. 9
Figure 6: Components of an Active Control System.................................................................14
Figure 7: Three-rodActuator scheme ..................................................................................... 16
Figure 8: ProposedCharlesRiver CrossingBridge ................................................................. 19
Figure 9: SAP2000 Model of Arch .......................................................................................... 23
Figure 10: Required Geometric Dimensionsfor Arch.............................................................. 24
Figure 11: Optimized Arch Shape ............................................................................................ 27
Figure 12: Moment DiagramofArch under Design Load Combination (1).............................. 28
Figure 13: Moment Diagramof Arch under CriticalLoad Combination (3)........................... 28
Figure 14: InteractionDiagramfor CriticalSection............................................................... 29
Figure 15: ProposedActuator Scheme on Arch........................................................................ 30
Figure 16: Graph of CriticalDesign Moments (Md) and Target Control Moment Distributions
(Mc *).............................................................................................................................................. 31
Figure 17: Graph of resulting ControlMoments (Mc) and sum of moments on structure (e)..... 32
Figure 18: InteractionDiagram of CriticalSection of Arch.................................................... 32

Page 5 of 50
MASSACHUSETTS
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
INSTITUTE OF TECHNOLOGY Page 5 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

1.0 Arch Structures

The first true arch structure was built approximately 44 centuries ago. However, it was
not until 44 centuries after the construction of the first arch that the mechanics of arches
was completely understood and formulated in the elastic theory of arches (Spofford,
1937). The enduring use of arches in civil engineering structures is attributed to the
aesthetic appeal and the structural efficiency of the system (Melbourne, 1995). The first
constructed arches were made of masonry. With masonry construction, the longest arch
structure spanned 300 feet. With the advent of more sophisticated materials, specifically
steel and reinforced concrete, arch spans have increased (Spofford, 1937). The New
River Gorges Bridge, is a 1700 feet metal arch bridge located in West Virginia; currently,
it is the longest constructed arch (Figurel). The longest concrete arch bridge is located in
Krk, Croatia and spans 1280 feet. The Natchez Bridge, is the longest concrete arch bridge
in the Unites States; it is located in Franklin, Tennessee and consists of two spans, with
the longer of the two arches spanning 582 feet (Figure2).

Figure 1: New River Gorges Bridge

TECHNOLOGY Page 6 of 50
MASSACHUSETTS
MASSACHUSETTS INSTITUTE OF
INSTITUTE OF TECHNOLOGY Page 6 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

Figure 2: Natchez Bridge

1.1 Types of Arches

There are three basic types of arches, a hingeless (fixed) arch, a two-hinged arch, and a
three-hinged arch. Older masonry arches have fixed supports at each base. The
corresponding six unknown reactions make fixed arches statically indeterminate
structures. The use of modern construction materials has made the possibility of structural
determinate arches a viable alternative. A three-hinged arch incorporates a hinge at the
crown of the structure in addition to hinges at the supports. The advantage of three hinged
arches is they are statically determinate and can be solved through equations developed by
the elastic theory; however, arches with fixed pinned supports are less stiff than structures
with fixed supports. A two-hinge arch, which is also statically determinate, incorporates
hinges at the supports, thereby preventing any moments from developing at the supports.
For the special case where the arch loading and configuration are perfectly symmetrical, all
types of arches can be also be solved through elastic equations. (Spofford, 1937)

A tied arch, incorporates an additional tension member that (typically) spans between the
abutments. This member resists the horizontal thrust, thus the foundation needs only to
resist vertical reactions. The tension tie must provide sufficient stiffness and prevent the
arch from displacing excessively.

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Page 7 of 50

DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

Most constructed arches consist of an arch spanning below the point of load application.
More recent arch structures have incorporated the arch above the point of load
application. The Oudry Mesly pedestrian bridge in Paris, France is an example of such a
structure (Figure 3). The rationale behind the placement of the arch is related to the
desired aesthetics and allowable clearances.

Figure 3: Oudry Mesly pedestrian bridge

1.2 Arch Theory

The distinguishing feature of an arch relative to other load carrying systems is that it
carries vertical loads primarily through axial action, and not flexure. Axial action is a
preferred load resisting mechanism since it makes the most efficient use of the
construction material. Figure 4 demonstrates a typical arch, including the pertinent
features. The arch shown in Figure 4 is a two-hinged arch. Governing force equilibrium
equations for arches are shown below Figure 5.

Page 8 of 50
MASSACHUSETTS
MASSACHUSETTS OF TECHNOLOGY
INSTITUTE OF
INSTITUTE TECHNOLOGY Page 8 of 50
-J

DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

lam(I

Figure 4: Typical Arch

V M

bX

Ry
Figure 5: Equilibrium Force Equations for an Arch

Nx = F sin # + V cos # Eqn. 1

Ny = F cos -V sin# Eqn. 2

I Fy = Ny - J b,dx + Ry Eqn. 3

ZFx = Nx + Jbxx - H Eqn. 4

E M = fbxdx + Jbxydy - Ryx - Hy Eqn. 5

As seen from the equations 1 and 2, the shear and axial forces within an arch are coupled.
As a consequence, under purely vertical load or purely horizontal load, both vertical and

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Page 9 of5

DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

horizontal reactions must be resisted at the foundation. This behavior is unlike typical
beams; a beam loaded with purely vertical load only resists vertical reactions at the
supports.

Although the arch equations allow for an arch to experience shear, bending, and axial
stresses, an optimized arch will experience primarily axial compression under load.
Specializing equation 5 for the case where no moments are allowed in the structure, and
assuming the arch is not subjected to any horizontal loads, the following equation results.

fby xdx - Ryx

y Jx)= H R Eqn.6

where y(x) represents the required distance between the springing (or tension tie) to the
centerline of the arch. The numerator in equation 6 is the exact expression for the
moment in a simply supported beam; the denominator is the value of the horizontal
reaction. From this equation, one can see that an optimized arch, i.e. one that experiences
no bending, is a scaled version of the moment diagram of a simply supported beam
subject to transverse loads.

In order for the optimized arch shape to be symmetrical, the applied loading must be
symmetrical. To attain a symmetric and parabolic arch curve, the design load must be a
uniformally applied vertical load loading throughout the arch. Fortunately, a series of
equal point loads equally spaced will also yield a symmetric shape that is nearly
parabolic; however, the resulting shape will not be a smooth curve. If the load on a
parabolic arch is not symmetric, or if the load pattern does not conform to the bending
diagram of the optimized shape, the arch will experience bending. The critical load
pattern for an arch is the case where only half the span is loaded. Under this condition the
maximum moment develops at the quarter point of the arch.

INSI IIU'JiE Ut' TCNL

IE(JHNULU(iYYF Page IU ot ~U
MASSACHUSETTS
MASSACHUSETTS INSTITUT Page 10 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

2.0 Funicular Method of Arch Design

For every load pattern on an arch, there exists an optimum arch shape, for which the
applied loading is resisted through purely axial action. No bending stresses exist in the
arch under this condition. There are several procedures used to determine the optimal
arch shape. All these procedures are based on the same objective: no moment in the arch.
For a given load configuration, there are an infinite number of optimal shapes. A
particular solution is generated by specifying a constraint, such as the arch length, three
pre-determined points along the arch, or the thrust force (Allen, 1998).

A funicular method of design, a graphical or numerical procedure, derived from equation

6, is used in this study to determine the optimum arch shape. The wordfunicular comes
from the Latin word, funiculus, meaning "string." (Allen, 1998) A funicular shape is the
form a piece of string would take under an applied load. Similar to a string, a funicular
shape resists load through axial action. A piece of string supporting uniform loads or
equally spaced concentrated loads takes the form of a parabola. Thus, a parabola is the
optimal shape for a structure resisting uniformly applied vertical loads through axial
action. (Heller, 1963).

The procedure, specialized for the constraint of three pre-determined points that the arch
must pass through, depends on the fact that the horizontal component of the force, Nx, at
each section of the arch is constant. This horizontal force is also equal to the thrust at the
foundation.

The first step in the funicular method is to determine the moment diagram (or
distribution) along the arch due to all vertically applied loads, including the self-weight of
the arch. The self-weight of the arch is typically not known during this first step. Thus,
for the first design iteration, a self-weight distribution is assumed. The moments are

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Page 11 of 50

DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

summed on a horizontal projection of the arch length. That is, the arch is assumed to be a
simply supported beam. To determine the value of the constant horizontal force, Nx, the
moment force at one constrained location is divided by the established distance between
the arch centerline and the springing. The required distance from the arch centerline to the
springing for every other point along the arch is found by dividing the moment at that
particular section by Nx. (Allen, 1998)

3.0 Disadvantages of Traditional Arch Design

Further study into the funicular method and equation 6 identifies an important
disadvantage of arch design: an arch shape is only optimizal for a particular load pattern.
If the distribution or magnitude of the loads are altered, the stresses along an arch change;
thus the arch shape will need to be altered for optimal conditions. However, arches are
static structures and it is unrealistic to change the structural geometry according to
transient load patterns. Consequently, traditional arch design utilizes the dominant load
configuration to determine the arch shape. For arch bridges, the dominant load (the
design load combination) is an arch loaded with full dead load and half the design live
load. For most other arches, the dominant load is an arch loaded with full dead load.
Under these conditions an arch will have purely axial action. It is usually under other load
cases, particularly live load that arches experience bending.

Depending on the arch structure, the bending due to secondary load cases may be
significant. The need to design an arch to resist bending in addition to axial loads
diminishes the attractiveness of arch structures. A more efficient arch design incorporates
a mechanism to limit the bending stresses in the arch to a pre-defined allowable limit.
Active controls applied to an arch provides an example of how efficiency of a structure
improves with the application of external control devices.

INS~'i FLUTE UP I ELHINULU~Y rage IL 01 DU

MASSACHUSETTS
MASSACHUSETTS INSTITUTEO CHNL Y Page 12 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

4.0 Active Control

A significant amount of inefficiency in structural systems results from the need to design
a structure for all possible design load conditions. Further, this type of structural design
is contingent upon always having reliable knowledge of the relative magnitude of all
loads the structure that will be applied to the structure (Connor, 1999 ).

The introduction of passive mechanisms of control has increased the efficiency of

structures by preventing loads from entering a structure, as in the case of base isolators
used to isolate a structure from seismic ground accelerations. Damping systems are also
utilized to dissipate the amount of energy, and therefore, stresses and movement the
structure experiences from dynamic loads. The shortcoming of these systems is that they
too are dependent on reliable knowledge of expected forces the structure will experience.
Both these devices are designed according to expected forces. (Connor, 1999)

The benefit of active controls is that they can accommodate unexpected loads imposed on
a structure. Active control devices are designed to monitor structural variables, such as
forces and displacements the structure is experiencing. The mechanism then determines if
the displacement or force exceeds pre-defined limits. In this case, the system determines a
set of actions that will change the structural state to an acceptable one. An actuator, a
mechanical device, induces a force or displacement according to the determined actions
to counteract externally applied forces or displacements (Connor, 1999).

4.1 Components of Active Control System

Active control devices are composed of three main components: the monitor, the
controller, and the actuator. The monitor is a set of sensors located along the structure

used to monitor structural variables discussed above. The controller is a cognitive

module, in some cases, with adaptive learning, which analyzes the sensor readings and

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Page 13 of 50

DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

uses a control algorithm to determine what actions to take to limit the forces or
displacements. The actuator implements decisions made by the controller. All the
components depend on an external energy source to investigate and implement decisions.
Figure 6 demonstrates the relationship and interaction between the components of an
active control system. (Connor, 1999 )

EXCITATION STRUCTURE RESPONSE

ACTUATOR

MO0NITOR MONITOR

to gure DEV:LopTHE ACTION PLAN ControlSStem

external f th C l as o omue
loading a T Wit tAhiknown response

| DECIDE ON COURSE OF ACTION

.|DENTIFY THE STATE OF SYSTEM -

CONTROLLER

Figure 6: Components of an Active Control System

Since force actuators require time to apply the force, the controller must have the

capability to predict forces a short distance off from the actuator. In the case of a bridge,
this scheme can be implemented if the controller was able to compute the weight and
velocity of a load moving along the deck. With this known information, the controller can
predict the impact of the load on the structure and send a signal to the actuator to apply a
counteracting force to maintain the forces in the arch to a prescribed limit. (Connor, 1999)

Current technology is able to support the sensors and algorithms required for the
controllers. However, actuator technology requires further advancement before active
control technology can be applied to large civil structures. Civil structures generally
require actuators to deliver large force to a system and to have short response times
(generally on the order of a few milliseconds). There are a number of different types of

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Page 14 of 50

DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

actuators that can impose a large force on a system; however, these actuators have long
response times. Most actuators require a large amount of energy to deliver the required
forces. Before active control technology can be implemented in civil structures, actuators
that are capable of delivering thousands of Newtons of force to a system within a few
milliseconds without requiring a large amount of external energy must be developed.
Alternately, the cost and ease of installing the system must improve so that multiple
actuators can be installed in a structure. (Connor, 1999)

4.2 Actuator Technology

There are three types of actuators technologies: hydraulic, electro-mechanical, and

electro-magnetic linear actuators. Each type of actuator has two components, a piston
and a gear mechanism. The gear mechanism applies a force to the piston in order to
translate the piston. Consequently, the piston applies a force to the structure and a
reaction force is felt be the gear mechanism. In cases where there is not adjacent surface
to absorb the reaction from gear mechanism, a self-equilibrating actuator system is
utilized. (Connor, 1999 )

An ideal actuator system is one that can deliver a large amount of force in a small
duration of time using minimal external energy. To control moment distributions on a
structure linear actuators must be place below or above a structure. One self-
equilibrating actuator scheme, applicable to the arch structure, is composed of three rods
and a gear mechanism (see Figure 7). (Connor, 1999)

INSTITUTE OF IEGHNULU(Y rage is at 50

MASSACHUSETTS INSTITUTE
MASSACHUSETTS OF TECY L Page 15 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

L
|+ - - - +

b)

A R

FL/4

Figure 7: Three-rod Actuator scheme

The vertically applied load F is equilibrated by vertical component of the force in the
diagonal rods. The forces cause a triangular moment field. A compressive force, of
magnitude F/2 is induced in the localized area. The shear force in the localized area is
also expected to increase. With a number of these actuator-rod configurations placed in
series, one can generate a piece-wise linear bending moment distribution.

4.3 Active Control Algorithm

The first step of determining the active control algorithm is to set a pre-defined limit on
the allowable stresses the system can withstand. For the case where moments need to be
controlled it is necessary to specify the maximum allowable moment, M(x)*.

The moment a structure feels at any point in time is the sum of the moment due to dead,
live, or other loads, Md, and the moment caused by the active control, M.

M* > M(x) = Md(x) + Mc(x) Eqn. 7

(91- TECNOOY
INSTITU'I E OF IELI-INULO(iY Page lb 01 iU
MASSACHUSETTS
MASSACHUSETTS INSTITUTE Page 16 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

The design objective is to limit M(x), the moment at any point in the structure, to M(x)*:
Using matrix notation, this simplifies equation (7) to the following:

- Md - M* < Mc & M * -Md Eqn. 8

From the range of acceptable control moments, Mc, one generates the target distribution,
Mc*. The sensor on an active control system senses the forces or moments at specified
locations on the structure. If the number of actuators on the system (r) is equal to the
number of sensor locations (n), the algorithm for the active control simplifies to applying
the actuator load to induce MC* at the sensor location. In this case, MC* will always be
equal to Mc. In most cases, the number of sensors on the system exceeds the number of
actuators (n>r). In this case, it is not possible to obtain the exact value of Mc* at every
sensor location. In this case, MC is a linear combination of moments caused by several
actuators. (Connor, 1999)

Mc ='Tm Eqn. 9

The error between Mc* and Mc is designated by e:

e=Mc*-Mc=Mc*-'Pm Eqn. 10

y is a vector of size nxr, which represents the distribution of moments due to a single
actuator; m represents the maximum magnitude of the moment field. In the case, where
the number of sensor exceeds the number of actuators, the goal of the control algorithm is
to minimize e using a least square algorithm. The least square method can be used to
determine an appropriate solution for e = 0. The error measure is taken as the norm of J
(Connor, 1999):

J=-eTe Eqn. 11
2

IELHNULUUY
OF TECNOOY
INSTITUTE OF 1'age 1/ at ~U
MASSACHUSETTS INSTITUTE
MASSACHUSETTS Page 17 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

Requiring J to be stationary with respect to each moment parameter contained in m leads

to a simple matrix equation

am =b Eqn. 12
am
where,

a= T T Eqn. 13

b=T Mc* Eqn. 14

5.0 Active-Controlled Arch Structure

Active control implemented on an arch structure is an ideal application of the

technology. With active controls, the arch shape and sections can be designed to carry
the dominant design load pattern through axial behavior. Rather than designing the arch
to resist stresses, particularly those produced by bending, caused by additional load cases,
other load cases will be supported by adjustments made be the active control mechanism.
The benefit of this system is an arch structure that primarily carries axial load and a
minimal amount of bending, regardless of the load configuration.

TECHNOLOGY Page 18 of 50
MASSACHUSETTS INSTITUTE OF
MASSACHUSETTS INSTITUTE OF TECHNOLOGY Page 18 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

6.0 Design of Proposed Charles River Crossing Bridge Arch

Figure 8: Proposed Charles River Crossing Bridge

6.1 Introduction
Components of the applicable design conditions and criteria of the proposed Charles River
Crossing Bridge (see Figure 8) will be utilized in this study. Active Control theory will be
implemented in the design procedure to demonstrate the reduction in bending stresses in
the arch under critical load conditions. The proposed bridge, a 750-foot span, segmental,
pre-cast concrete, two-hinge arch bridge is seen on Figure 8. The rise of the proposed
bridge is 250 feet from the waterline. A unique feature of this bridge is the location of the
tie. Typically, the tie on an arch is located at the foundation level. Since the bridge spans
over a water crossing, the tie had to be placed at higher elevation. Currently, the tie is at
deck level and braced by the deck superstructure. The arch is constructed of concrete box
sections, which taper in depth from the crown of the arch to the base. Loads are imposed
on the arch through a series of suspension cables. For the design, a two-dimensional
model of one of the arches is idealized. For purposes of demonstrating the use of active
controls to minimize bending in the arch, the out of plane forces caused by the inclined
cables are neglected.

Page 19 of 50
MASSACHUSETTS
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
INSTITUTE OF TECHNOLOGY Page 19 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

The design of the arch is a two step procedure. The first phase involves determining the
optimal shape of the arch for the dominant design load. Further, the arch must be checked
to ensure it provides adequate axial capacity for all load combinations. The second phase
of the design is the implementation of the active control system.

The design load combination for an arch is full dead load and half the live load. The dead
weight is typically significant portion of the dead weight. The arch cross-section is
optimized for the required stress capacity for the design load combination. The iterative
nature of arch design is a result of changes in the arch self-weight as the arch cross-
section and shape is revised. The basic design procedure utilized in this case to determine
the optimum arch shape is outlined in the following steps.

1. Determine imposed loads on the arch, specifically imposed dead and live loads.
2. Specify geometric constraints, specifically three pre-defined points.
3. Determine funicular shape of arch for dead and live loads from superstructure.
4. From estimated value of axial force near crown, establish required section properties
required at arch crown.
5. Using coordinates found from funicular method, perform regression analysis to
determine fourth-degree polynomial of arch shape
6. Estimate weight of arch per linear foot (utilizing derivative of polynomial)
7. Determine funicular shape for dominant load case.
8. Verify adequacy of sections for design axial forces for all load combinations.
9. Iterate section properties, as necessary
10. Return to step 4, as necessary.

6.2 Loads

TECHNOLOGY
OF TECHNOLOGY
INSTITUTE OF Page 20 ot 50
MASSACHUSETTS
MASSACHUSETTS INSTITUTE Page 20 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

There are two distinct types of loads imposed on an arch: loads from the superstructure
and the self-weight of the arch. Superstructure loads include design dead and live loads on
the superstructure. The superstructure loads are transmitted to the arch through the cables.

6.2.1 Superstructure Loads

Table 1 lists the imposed dead and live loads from the bridge superstructure. The ID
number listed on Table 1 corresponds to points found on Figure 9. Typically each cable
is spaced at 25.8 feet along the arch axis. Under the dominant design load pattern each
cable carries approximately 280 kips of load. From Table 1 one can see the loads in the
cables change at the extreme ends of the arch. Geometric compatibility between the arch,
the deck, and required pedestrians access forced the re-alignment of the cables at these
locations. Consequently, cables at the ends have a greater tributary area of load to carry.

TECHNOLOGY
OF TECHNOLOGY Page 21 of 50
MASSACHUSETTS INSTITUTE OF
MASSACHUSETTS INSTITUTE Page 21 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

Table 1: Imposed Design Dead and Live Loads

ID x-Distance Cable Dead 50% Cable Total
Load Live Load Imposed
Load

feet kips kips kips

0.0
3 45.8 399 42.69 441
4 71.5 148 51.54 199
5 97.2 243 49.59 293
6 122.9 220 50.11 270
7 148.6 228 50 278
8 174.3 228 50 278
9 200.0 229 50 279
10 225.7 230 50 280
11 251.4 231 50 281
12 277.2 232 50 282
13 302.9 232 50 282
14 328.6 232 50 282
15 354.3 233 50 283
16 367.2
17 380.0 233 50 283
18 405.7 232 50 282
19 431.4 232 50 282
20 457.1 232 50 282
21 482.8 231 50 281
22 508.6 230 50 280
23 534.3 229 50 279
24 560.0 228 50 278
25 585.7 227 50 277
26 611.4 224 50 274
27 637.1 227 50 277
28 662.8 210 50 260
29 688.5 260 50 310
716.3

TECHNOLOGY
OF TECHNOLOGY
INSTITUTE OF Page 22 of 50
MASSACHUSETTS INSTITUTE
MASSACHUSETTS Page 22 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

Figure 9: SAP2000 Model of Arch

6.2.2 Self-weight

The arch was designed to taper in depth from a square box section at the crown to
rectangular box section at the base. The taper is implemented for structural reasons: the
compressive stresses in the arch increase towards the supports. The taper allows for the
area of the section to increase in proportion to the increase in axial stresses. The variation
in depth and the calculation of the arch self-weight was formulated according to Melan's
method of arch design. The method approximates the change in depth, d, according to
the secant of the enclosed angle (Melan, 1915):
d= dsec# Eqn. 15

do = depth of arch at crown

# = enclosed angle

Accordingly, the self-weight, q, of the arch is expressed as:

q = dywsec#, Eqn. 16

or q = wdo(1 + tan 2 ) Eqn. 17

w = width of arch

y= unit weight of concrete

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Page 23 of 50

DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

150 pcf

6.3 Funicular method

The funicular method as described in Section 2.0 is used in this study to determine the
optimal arch shape. Specifically, the numerical version of the procedure specialized for
the case where three pre-defined points along the arch are known is utilized.

6.3.1 Determination of Constraints

As discussed earlier, the deck braces the tension tie of the arch. The deck in turn must
connect to existing viaducts on either end of the bridge. These geometric constraints
dictated the location of 2 of the three pre-defined points required for the funicular
method. The arch must intersect with the deck at the end of the deck at both ends. The
third required point was determined by the desired maximum height of the arch, 250 feet,
above the waterline. Figure 10 describes each of these points.

696' - 4"
Figure 10: Required Geometric Dimensions for Arch

Page 24 of 50
MASSACHUSETTS OF TECHNOLOGY
INSTITUTE OF
MASSACHUSETTS INSTITUTE TECHNOLOGY Page 24 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

6.3.2 Determination of optimal arch shape

The first iteration of the design was the determination of the funicular arch shape from
only the applied superstructure loads using the method described in Section 2.0. A
regression analysis was performed on the resulting coordinates to determine a fourth
degree polynomial, h(x), describing the arch shape. The polynomial was necessary to
approximate the tangent of the enclosed angle with the first derivative of the arch shape:
d
tan$ - dh(x)
dx

The first iteration through the funicular method revealed an approximate value of the
axial force in the arch near the crown. Since the first iteration was performed using only
the superstructure loads, the computed axial force was significantly less than an iteration
including the self-weight of the arch. The initial value for the arch cross-section was
determined by assuming the required axial capacity of the arch at the crown was double
the value of the axial force found from this first iteration.

All further iterations of the funicular method utilized the dominant load case, full dead
load plus half the design load, to determine the optimal arch shape. The calculations for
the funicular arch shape are found in Appendix 1. The values in the chart are from the
final iteration. The procedure utilized is as described in Section 2.0.

6.4 Computer Model

After a couple of iterations through the funicular method, a model of the arch was
constructed in SAP2000. Figure 9 depicts the SAP2000 model. Since SAP2000 is not
able to idealize a tapered member, sections of the arch were discretized. The depth of
each discretized section was found by maintaining the weight of each section. The weight

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Page 25 of 50

DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

of each segment was found using the eqn 16, which describes the weight per linear foot
of the arch. Appendix 2 contains a spreadsheet of the calculated discretized depths.
SAP2000 was used in conjunction with interaction diagrams (see Appendix 4) to ensure
the axial capacity of the arch was sufficient to resist loads resulting from the governing
load cases. Following are the three governing load cases:

Load Combination 1 (design load pattern): Dead Load plus half the Live Load over the
entire arch.
Load Combination 2: Dead Load plus the entire Live Load over the entire arch.
Load Combination 3 (critical load pattern): The arch loaded with full Dead Load, but
only half the arch loaded with Live Load.

In total, four iterations were performed until convergence of the stresses and the
coordinates was observed.

OF TECHNOLOGY
INSTITUTE OF Page 26 ot 50
MASSACHUSETTS INSTITUTE
MASSACHUSETTS TECHNOLOGY Page 26 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

6.5 Optimized Arch Shape

250.00

200.00
-A
150.00

100.00

N
50.00

0.00
0
i
. 100.00 200.00
.,
30a.00 40a.00 500.00 600.00 70C.00 80C .00

-50.00
x -distance (ft.)

* points from funicular method

* ideal parabola
- Poly. (points from funicular method)
-Poly. (ideal parabola)

Figure 11: Optimized Arch Shape

A graph of the final arch shape is found on Figure 11. Also shown is a graph of an
idealized parabolic shape of the arch found using the pre-defined geometric points. As
shown, the points do not differ significantly, but the small variation eliminates bending
from the arch. The base of the arch shown in Figure 11 is actually the intersection of the
arch with the deck. During the design, it was realized that continuing the arch curve to the
foundation level induced major bending in the arch. The additional bending force is
proportionate to the vertical reaction at the deck level and the horizontal eccentricity
between the arch at the deck level and the foundation. Eliminating this eccentricity
completely eliminates the additional bending stress. Thus, to minimize the amount of
bending in the arch, the arch curvature should discontinue at the tie level. Instead, vertical
supports should extend from the arch at the deck level to the foundation.

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Page 27 of 50

DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

6.6 Discussion of Design Stresses

The last load combination is the critical load combination for arches since the non-
symmetrical loading maximizes the bending stresses in the arch, as seen from Figures 12
and 13. The bending stresses in the arch under Load Combination one are negligible
(Figure 12). The moments observed in the arch under Load Combination one is due to the
discretized sections. The moments in the arch under Load Combination three are
significant (Figure 13). The moment is maximized at the quarter point where the moment
is approximately 16,800 kip-ft. With traditional arch design, the arch cross-sections are
designed to resist these moments.

Figure 12: Moment Diagram of Arch under Design Load Combination (1)

Maxumen Unfactored 'I

Moment: 16,800 kip-ft

Figure 13: Moment Diagram of Arch under Critical Load Combination (3)
The interaction diagram for the section where the moment is maximized is shown in
Figure 14. At this location, the section is approximately 9.5 ft. deep. The section is
designed to have adequate axial capacity for all governing load combinations. However,
as traditional design would suggest, the section was not designed for moment capacity for

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Page 28 of 50

DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

all load combinations. As the interaction diagram in Figure 14 demonstrates, the section
located at the point of maximum moment does not have sufficient moment capacity to
resist the force under the critical load combination. The use of active control on the arch
structure eliminates the need to design sections for critical moment stresses.

14000
12000
10000
8000
6000

a. 4000
2000
0
-2000
-4000
*M, kip -ft

-- + critical section

Figure 14: Interaction Diagram for Critical Section

7.0 Design of Active Control System

The objective of the active controls in the arch is to limit the moments in the arch to a
pre-defined limit. The three-rod actuator scheme discussed in Section 4.2 is used on the
arch to control moments. The benefit of this actuator scheme is that the forces and
moments generated in the arch are restricted to a localized area. Further, the stresses
induced by a particular actuator are equilibrated within the restricted length.

7.1 Configuration of Actuators

The actuators are placed in series along the arch in order to generate a piece-wise linear
moment field. The actuator configuration is as shown in Figure 15.

MA~ALtiUSt.1 IS I1NS1IIUIJ~. U1 ItCHNULUUY Page 29 ot 50

MASSACHUSETTIS INS IUEO CH LG Y Page 29 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

n.

L r3

Figure 15: Proposed Actuator Scheme on Arch

For purposes of implementing the actuator scheme, the arch was idealized by a series of
twenty feet long beams placed in series. Moments are monitored along the arch at the
start, midpoint, and end of each beam segment. The locations where moments are
monitored are designated by n, (Refer to Figure 15). The piston corresponding to each
actuator is placed at the intersection of every beam segment. The location of an actuator
is designated in Figure 15 by ri. The moment field produced by each actuator is
distributed over the lengths of two beams segments. The actuators are placed normal to
the intrados (interior) of the arch. Since each moment field is restricted to a relatively
short length of the arch, the length over which the moment field is applied can be
approximated to be linear. Consequently, traditional beam theory can be used to
determine resulting shear, axial and moments over the localized length.

7.2 Active Control Algorithm

In order to demonstrate the effect of active control on an arch, M*, the limiting moment
capacity, was a constant value for all areas along the arch. Since the arch tapers, a more
realistic, and more complex, algorithm would include varying M* along the length of the
arch. The value of M* at each point would be selected according to the bending capacity
of the arch section and the axial load the section is experiencing. For this study, M* was
selected to be 6000 kip-ft. This value is slightly less than the maximum allowable stress
in the arch crown with the current axial stress in the arch.

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Page 30 of 50

DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

A Matlab program was formulated to compute the actuator forces required to limit the
moments throughout the arch to M*. A copy of the Matlab input is seen in Appendix 3.
The Matlab algorithm was designed to find the appropriate control moments when the
arch is under the critical load pattern.

7.3 Results of Active Control Algorithm

Below are the results from the Matlab algorithm. Figure 16 demonstrates the applied
loading, Md, the range of acceptable control moments, Mc- and Mc+, and the target
control moment, Mc*.

30000
20000
A-
10000 Md
------ Mc-
0 ------ Mc+
-10000 S200 /4008 00 Mc*

-20000
-30000
Distance along Arch Circumference, ft.

Figure 16: Graph of Critical Design Moments (Md) and Target Control Moment Distributions
(Mc*)

The area between the dotted curves of Mc- and Mc+ represent the range of acceptable
control moments. The target moment field, Mc*, was selected based on minimizing the
required control moment to bring the arch within the acceptable range of bending
stresses. Figure 17 depicts the target control moment and the resultant control moments
found from Matlab. Also included is e, the summation of the design moments and the
control moments. As seen in the Figure, the resultant control moments, Mc, are very

TECHNOLOGY
OF TECHNOLOGY Page 31 of 50
MASSACHUSETTS
MASSACHUSETTS INSTITUTE OF
INSTITUTE Page 31 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

close to the target moment field, Mc*. The graph of e demonstrates that maximum
moment along the arch is approximately 6000 kip-ft. This was the expected result.

15000

10000
- Mc*
5000
iC
0 0
0 200 /40\ 60/ 800 -- e
2 -5000

-10000 P v-I
Distance along Arch Circumference, ft.

Figure 17: Graph of resulting Control Moments (Mc) and sum of moments on structure (e)

The actuator force required to deliver the peak control moment, which coincides with the
location of the critical section described above, is 2200 kips. Using the same interaction
diagram created above, the stresses at the critical section are compared before and after
the implementation of the active control system (see Figure 18)
14000 -

-2000.0 100 1 2000 2

O-n, klP-ft.
.. critical bending stress (traditional design)
-.-.. wth actbe control

Figure 18: Interaction Diagram of Critical Section of Arch

OF TECHNOLOGY Page 32 of 50
MASSACHUSETTS INSTITUTE OF
MASSACHUSETTS INSTITUTE TECHNOLOGY Page 32 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

As the interaction graph demonstrates, the section is now sufficient to resist required
stresses even at the critical loading. The moment in the section decreased by 54%.

MASSA(JHUSETTS INSTITUTE OF TECHNOLOGY Page 33 of 50

MASSACHUSETTS INSTTT OYGFTC Page 33 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM LOADS

8.0 Conclusion

Arches have been used for centuries in structural engineering. The continuing use of
arches is due to the efficiency of the system and the aesthetic appeal of the structure.
Arch structures resist load primarily through axial action. The weakness of an arch is it
may develop considerable bending stresses when subject to non-uniform or random load
patterns. The use of active controls in an arch successfully decrease the bending stresses
in an arch to a pre-scribed limit. The result is an arch structure, which resists all load
patterns primarily through axial action. Consequently, the construction material is utilized
very efficiently.

This result is not without consequence or cost. The actuator requires external energy. The
cost of material, labor and construction of the traditional method must balance the cost of
the actuator scheme, including the physical components and the energy source. A broader
analysis should also incorporate long term savings of implementation of the actuator as a
means to take care of future load requirements, the aesthetic appeal of a less bulky
structure, the cost of maintaining the system, and the tested reliability of such a system.

Currently, an analysis of the system with current technology would probably suggest the
traditional design. In the future, when the technology able to support such a system at a
economical cost is available and scientific and pubic trust in such a system has been
established, the use of active controls to build more reliable, intelligent, civil systems will
be a viable alternative.

Page 34 of 50
MASSACHUSETTS OF TECHNOLOGY
INSTITUTE OF
MASSACHUSETTS INSTITUTE TECHNOLOGY Page 34 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS

APPENDIX 1

Typical cable spacing 25.716

ID x-Distance Cable 50% Total

Dead Cable Imposed
Load Live Load Load

feet kips kips LOAD

1 0.0
2 45.8 399 42.69 441
3 71.5 148 51.54 199
4 97.2 243 49.59 293
5 122.9 220 50.11 270
6 148.6 228 50 278
7 174.3 228 50 278
8 200.0 229 50 279
9 225.7 230 50 280
10 251.4 231 50 281
11 277.2 232 50 282
12 302.9 232 50 282
13 328.6 232 50 282
14 354.3 233 50 283
15 367.2
16 380.0 233 50 283
17 405.7 232 50 282
18 431.4 232 50 282
19 457.1 232 50 282
20 482.8 231 50 281
21 508.6 230 50 280
22 534.3 229 50 279
23 560.0 228 50 278
24 585.7 227 50 277
25 611.4 224 50 274
26 637.1 227 50 277
27 662.8 210 50 260
28 688.5 260 50 310
29 716.3 1 1

Page 35 ot 50
MASSACHUSETTS OF TECHNOLOGY
INSTITUTE OF
MASSACHUSETTS INSTITUTE TECHNOLOGY Page 35 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS

APPENDIX 1

Moment at
Section
ID Reaction due to due to self-
calcu external weight
loads

kip-ft kip-ft

1 0 0
2 295933 166271 87664
3 128539 248331 128779
4 181106 325266 164747
5 160222 394677 196045
6 157827 472524 223083
7 150467 512462 246216
8 144212 560640 265744
9 137589 601635 281920
10 130740 635416 294952
11 123759 661964 305007
12 116649 681265 312214
13 109475 693309 316662
14 102259 698092 318409
15 697126 318275
16 94996 695610 317476
17 87695 685863 313850
18 80375 668854 307484
19 73035 644590 298298
20 65656 613077 286172
21 58262 574332 270953
22 50843 528374 252444
23 43394 475232 230407
24 36156 414950 204558
25 28764 347548 174560
26 21891 273093 140024
27 13881 191526 100497
28 8606 103282 55461
29 0 0

Reaction 3633
atA
Reaction 3741
at B

Page 36 of 50
MASSACHUSETTS OF TECHNOLOGY
INSTITUTE OF
MASSACHUSETTS INSTITUTE TECHNOLOGY Page 36 of 5
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS

APPENDIX 1

Z -height at center 210.5 ft. from base

Horizontal Force 4823.757 kips

COORDINATES FOUND FROM RESULTING FOURTH DEGREE POLYNOMIAL

FUNICULAR METHOD y = -7.654E-10x4 + 1.106E-06x3 - 2.145E-03x2
+ 1.250E+00x - 1.543E-01

x y x z

25 2977
71.47 781 45.76 52.66
97.19' T01.5R 71.47 78.62
122D 12246 97.19 102.01
122.90 122.95
17M. 15.28 1486 141.49'
2003 171.32' 174.32 -- 157.7 1
22.7 1. 7 200.03 171.68
225.7--18W3.4
~2777 2046 251.5 193.06
4. 20.25
32.87 205.m 42771 202.5
38 209.7 482.87 256
5T29 210.73' 528.55 27.0
36722 210.50 54.269 2.71
38 210.02 36T2 210.47'

431.2 2024 4571 2070

45.1 195.47 1.2 202.18
48.4 16. 559.97 129461
508.55 175.23 637.1 185.2
53426 ~161.87 662.81 165
559.7 146.28 53426 161.72
585.6 1284 ~559.7 146.1
585.6 128.30

68. 32.9
71627' 0.0 688.52 32.62
700 19.38

Page 37 of 50
MASSACHUSETTS
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
INSTITUTE OF TECHNOLOGY Page 37 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS

APPENDIX 2

Section Length Weight d

TD ft kips ft.

1 38.9 246.2 11.1

2 30.9 186.3 1.
3 36.5 210.3 9.8 q = weight per linear foot
4 34.8 190.5 9.2 q=-.522E-17x 7+. 132E-13x6 -.294E-1 0x 5+.358E-7x 4
5 33.2 173.3 8.6 -3.56E-5x3+.0216x2-10.368x
6 31.7 158. 8
7 30.4 145.7 7.7 Depth Calculation for Discretizied SAP Model
8 29.3 135 7
9 28.3 126 6 W=L*A*y L= length of section
10 27.4 118.8 6.6 W= L*( 6dn - 3(dn-3))*y y = unit weight of concrete
11 26.8 113.1 6.4
12 26.3 108.8 62 A = Area of section
13 25.9 105.9 6.1 = 6dn - 3(dn-3)
14 25.7 104.4 6.0 dn= depth of section
15 12.9 52.4 6.0
16 12.8 51.9 6.0
17 25.9 105.4 6.1 dn= W
18 26.2 107.9 6.1 3Ly
19 26.6 111.8 6.3
20 27.3 117. 6.5
21 28.0 123.9 6.8
22 29.0 132.3 7.1
23 30.1 142.4 .5
24 31.3 154.5 8.0
25 32.7 168.6 8.5
26 34.2 184.9 9.0
27 36.0 203.7 9.6
28 37.8 225.3 10.2
29 17.5 108.3 10.7
30 25.6 162.5 11.1

Total Weight 4275.6

Page 38 of 50
MASSACHUSETTS
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
INSTITUTE OF TEH OY Page 38 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS

APPENDIX 3

clear

diary thesisfinal.diary

hold on

%Units are kip, feet

% ARCH COORDINATES (x,y);

% X contains a list of observed points (n)

X=wkiread('xvalues');

% p is the fourth degree polynomial representing the arch curve

p =[-7.654e-10 1.106e-6 -2.145e-3 1.25 -1.543e-1] ;

% TO DETERMINE Y COORDINATES, EVALUATE THE POLYNOMIAL AT X

Y = polyval(p,X);

% PLOT ARCH COORDINATES

%plot (X, Y);

% Determine length of arch segments (all segments should be 20 ft. in

length)

for n=1:2:(length(X)-1)
for i=1:((length(X)-1)/2)
L(i)=sqrt((X(n+2)-X(n)).^2+(Y(n+2)-Y(n) .^2)
end
end
L

% Md is a vector of the moments the arch is experiencing under the

critical load

Md=wklread('moments')

% Mstar is a vector of the allowable moments

Mstar = 6000;

%Determine range of Mc, the control Moment

for j=1:length(Md)
Mc1(j)=Mstar-Md(j);
end;

Mci=Mci'

Page 39 at 50
MASSACHUSETTS
MASSACHUSETTS OF TECHNOLOGY
INSTITUTE OF
INSTITUTE TECHNOLOGY Page 39 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS

APPENDIX 3

for j=l:length(Md)
Mc2(j)=-Mstar-Md(j);
end;

Mc2=Mc2'

%Determine the target, optimal control moment

for j=1:length(Mcl)
if Mcl(j)>O
if Mc2(j)>0;
Mt(j)=Mc2(j);
end;
end;
if Mcl(j)<0;
if Mc2(j)<O;
Mt(j)=Mcl(j);
end;
end;
if Mcl(j)>0;
if Mc2(j)<O;
Mt(j)=O;
end
end
if Mcl(j)<0;
if Mc2(j)>O;
Mt (j)=O;
end
end
end;

Mt=Mt'

% TYPICAL PSI FUNCTION

TYP = [0 .75 1 .75 0]'

psi=zeros(2*length(L)+1,length(L)-1);

for i=1:length(L)-1
psi((2*i-1):(2*i+3),i:i)=TYP(1:5,1:1);
end;
psi;

%The solution to the unknown moments is aM=b, where M=inverse(a)*b

a=psi' *psi;

b=psi'* (Mt);

M=inverse(a)*b

Mc=psi*M

e=Md+Mc

for i=1:length(e)

Page 40 of 50
MASSACHUSETTS
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
INSTITUTE OF TECHNOLOGY Page 40 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS

APPENDIX 3

if e (i)>0
error(i)=e(i)-Mstar;
else
error(i)=e(i)+Mstar;
end;
end;

error=error'

plot (X, Md, 'y')

plot(X,Mcl, 'm')
plot (X,Mc2)
plot(X,Mt, 'r')
%plot(X,M, 'g')
plot (X, e, 'b')
plot (X, Mc, ' k')
%plot(X,error, 'm')

wklwrite('outputx',X)
wklwrite('outputmcl',Mcl)
wklwrite('outputmc2',Mc2)
wklwrite('outputmt',Mt)
wklwrite('outputmd',Md)
wklwrite('outputmc',Mc)
wklwrite('outpute',e)

Page 41 of 50
MASSACHUSETTS
MASSACHUSETTS OF TECHNOLOGY
INSTITUTE OF
INSTITUTE TECHNOLOGY Page 41 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS

APPENDIX 4
MATERIAL PROPERTIES

fc 4500 psi
fy 60 ksi

SECTION TYPES

Section 6x6 6 x 9.5 6x11

height 72 in 90 in 132 in
base 72 in 72 in 72 in
tw 18.00 in 18.00 in 18.00 in
tf 18.00 in 18.00 in 18.00 in

SECTION PROPERTIES

Section 6x6 6 x 9.5 6x11

Ixx 101 ft4 188 ft4 538 ft4

lyy 101 ft4 125 ft 4
180 ft4
Sxx 17 ft3 25 ft3 49 ft3
Syy 17 ft3 21 ft3 30 ft3
Asx 18 ft 2 23 ft2 33 ft 2
Asy 18 ft2 18 ft2 18 ft 2
Ag 27 ft2 32 ft2 42 ft 2

TECHNOLOGY
INSTITUTE OFFTCNL
MASSACHUSETTS INSTTT Page 42 of 50
MASSACHUSETTS Y Page 42 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS
APPENDIX 4

REINFORCEMENT

Section 6x6 6 x 9.5 6x11

Layer 1
Bar # 9 9 10
No. of Bars 15 15 15
Steel Areal 14.91 in2 14.91 in2 18.41 in2
Depthl 67 in 85 in 127 in

Layer 2
Bar # 9 9 10
No. of Bars 10 15 10
Steel Area2 9.94 in2 14.91 in2 12.27 in2
Depth2 62 in 80 in 122 in

Layer 3
Bar # 9 9 10
No. of Bars 15 15 15
Steel Area3 14.91 in2 14.91 in2 18.41 in2
Depth3 5 in 5 in 5 in

Layer 4
Bar # 9 9 10
No. of Bars 0 10 10
Steel Area4 0 in2 9.940195505 in2 12.2718463 in2
Depth4 10 in 10 in 10 in

p
1.02% 1.21% 1.01%

p
m in. 0.34% 0.32% 0.32%

p-max

Page 43 of 50
MASSACHUSETTS OF TECHNOLOGY
INSTITUTE OF
MASSACHUSETTS INSTITUTE TECHNOLOGY Page 43 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS
APPENDIX 4

MOMENT CAPACITY CALCULATIONS (ref. ACI 10.2.7)

Balanced Moment

p 0.825
Su 0.003
fy 60 ksi
E 29000 ksi
0.002069
(bend.) 0.7

Section 6x6 6 x 9.5 6x11

c(bal) 39.65 in 50.31 in 75.16 in

a 32.71 in 41.50 in 62.01 in
.85*f'c 3825.00 psi 3825.00 psi 3825.00 psi
Cc -6983 k -8194 k -11017 k
Rc 16.45 20.90 32.06
Mc -162059 k-in -240968 k-in -474890 k-in

F1 0.00207 0.00207 0.00207

(71 60.0 60.0 60.0
T1 895 k 895 k 1104 k
M1 24465 k-in 31038 k-in 57252 k-in

82 0.0017 0.0018 0.0019

u2 49.0 51.4 54.2
T2 487 k 766 k 665 k
M2 10891 k-in 30893 k-in 54784 k-in

F-3
-0.0026 -0.0027 -0.0028
u3 -60.0 -60.0 -60.0
T3 -895 k -895 k -1104 k
M3 -31001 k-in -31001 k-in -38273 k-in

4 -0.0022 -0.0024 -0.0026

ur4 -60.0 -60.0 -60.0
T4 0 k -596 k -736 k
M4 0 k-in -17685 k-in -21834 k-in

phi 0.70 0.70 0.70

N -6496 k -8024 k -11088 k

-4547 k -5617 k -7762 k
Mn 19035 k-ft 29299 k-ft 53919 k-ft
-Mn 13324 k-ft 20509 k-ft 37744 k-ft

UP 1EC~HNULU(jY
INSTITUIE OF
MASSACHUSETTS INSTITUTE
rage ~'r4 01 ~U
MASSACHUSETTS HNLO Y Page 44 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS
APPENDIX 4

Pont 1

Section 6x6 6 x 9.5 6x11

c_1 72.00 in 90.00 in 132.00 in

a 59.40 in 74.25 in 108.90 in
.85*f'c 3825.00 psi 3825.00 psi 3825.00 psi
Cc -11402 k -13013 k -16772 k
Rc 29.70 37.13 54.45
Mc -338626.3 k-in -483094.6 k-in -913227.8 k-in

El -0.0002 -0.0002 -0.0001

al -6.0 -4.8 -3.3
TI -90 k -72 k -61 k
M1I -450 k-in -360 k-in -303 k-in

E2 -0.0004 -0.0003 -0.0002

a2 -12.1 -9.7 -6.6
T2 -120 k -144 k -81 k
M2 -1201 k-in -1441 k-in -809 k-in

E3 -0.0028 -0.0028 -0.0029

a3 -60.0 -60.0 -60.0
T3 -895 k -895 k -1104 k
M3 -59939 k-in -76042 k-in -140267 k-in

F4 -0.0026 -0.0027 -0.0028

a4 -60.0 -60.0 -60.0
T4 0k -596 k -736 k
M4 0 k-in -47713 k-in -89830 k-in

phi 0.70 0.70 0.70

N -12296 k -14504 k -18613 k

-8607 -10153 -13029
Mn 9374 k-ft 15368 k-ft 28830 k-ft
*-Mn 6562 k-ft 10758 k-ft 20181 k-ft

Page 45 of 50
MASSACHUSETTS
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
INSTITUTE OF TECHNOLOGY Page 45 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS
APPENDIX 4

Point 2

Section 6x6 6 x 9.5 6 x 11

c_2 10.00 in 10.00 in 12.00 in

a 8.25 in 8.25 in 9.90 in
.85*f'c 3825.00 psi 3825.00 psi 3825.00 psi
Cc -2272 k -2272 k -2726 k
Rc 5.00 5.00 6.00
Mc -11360 k-in -11360 k-in -16359 k-in

E, 0.0171 0.0225 0.0288

al 60.0 60.0 60.0
TI 895 k 895 k 1104 k
Ml 50993 k-in 67096 k-in 127014 k-in

E2 0.0156 0.0210 0.0275

u2 60.0 60.0 60.0
T2 596 k 895 k 736 k
M2 31013 k-in 62623 k-in 80994 k-in

E3 -0.0015 -0.0015 -0.0018

a3 -43.5 -43.5 -50.8
T3 -649 k -649 k -934 k
M3 -3243 k-in -3243 k-in -6539 k-in

E4 0.0000 0.0000 -0.0005

c74 0.0 0.0 -14.5
T4 0.00 k 0.00 k -178 k
M4 0.0 k-in 0.0 k-in -356 k-in

phi 0.90 0.90 0.90

N -1430 k -1131 k -1998 k

-1287 -1018 -1798
Mn 8051 k-ft 12027 k-ft 19272 k-ft
-Mn 7246 k-ft 10824 k-ft 17345 k-ft

OF TECHNOLOGY
INSTITUTE OF Page 46 of 50
MASSACHUSETTS INSTITUTE
MASSACHUSETTS TECHNOLOGY Page 46 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS

APPENDIX 4

Pure Bending

Section 6x6 9.5 x 14 9.5 x 18

c(bend) 5.79 in 7.38 in 7.43 in

a 4.77 in 6.09 in 6.13 in
.85*f'c 3825.00 psi 3825.00 psi 3825.00 psi
Cc -1315 k -1677 k -1687 k
Rc 2.89 3.69 3.71
Mc -3804 k-in -6191 k-in -6266 k-in

E1 0.0317 0.0315 0.0483

al 60.0 60.0 60.0
TI 895 k 895 k 1104 k
Ml 54763 k-in 69438 k-in 132065 k-in

82 0.0291 0.0295 0.0463

a2 60.0 60.0 60.0
T2 596 k 895 k 736 k
M2 33526 k-in 64965 k-in 84362 k-in

E3 -0.0004 -0.0010 -0.0010

-11.8 -28.1 -28.4
T3 -176 k -419 k -523 k
M3 -139 k-in -997 k-in -1270 k-in

0.0022 0.0011 0.0010

a4 60.0 30.9 30.1
T4 0 k 307 k 370 k
M4 0 k-in 803 k-in 952 k-in

N Ok Ok Ok
Mn 7686 k-ft 11866 k-ft 18743 k-ft
-Mn 5380 k-ft 8306 k-ft 13120 k-ft

TECHNOLOGY Page 47 ot 50
MASSACHUSETTS INSTITUTE OF
MASSACHUSETTS INSTITUTE OF TECHNOLOGY Page 47 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS
APPENDIX 4

CAPACITY CALCULATIONS (cont.)

AXIAL COMPRESSION CAPACITY (ref. ACI 10.3.5)

Section 6x6 6 x 9.5 6x11

*-comp 0.7

Ag 3888 in2 4536 in2 6048 in2

2
As 39.76 in 54.67 in2 61.36 in2

Pn 15470 k 18517 24036

10829 k 12962 k 16825 k
4 -Pn

AXIAL TENISON CAPACITY

-tens 0.9

Pn 2385.65 3280.26 3681.55

*-Pn 2147.08 2952.24 3313.40

Page 48 0± 50
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
INSTITUTE OF TECHNOLOGY Page 48 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS

APPENDIX 4
12000

10000

8000

6000
U)

0. 4000

2000

-2000

-4000

-- M , kip-ft.
-- 6-ft. x 6-ft. section --a- Force at Crown under Critical load

18000
16000
14000
12000
10000
U) 8000
-e
6000
(L 4000
2000
0
-2000
-4000
-6000

0-Mn, kip-ft.

- 6-ft. x 11-ft. section

MASSALHUS~ 115 INSTITUTE OF TECHNOLOGY Page 49 of 50

MASSACHUSETTS INSTTT TEHO YF Page 49 of 50
DESIGN OPTIMIZATION OF PARABOLIC ARCHES SUBJECT TO NON-UNIFORM
LOADS
APPENDIX 4

References
Allen, Edward and Zalewski, Waclaw. Shaping Structures, Statics. New York,
New York: John Wiley and Sons, 1998
Connor, Jerome and Klink, B. Introduction to Structural Motion Control.
Cambridge, MA: MIT, 1999
Heller, Robert and Salvadori, Mario. Structure in Architecture. Englewood Cliffs,
New Jersey: Prentice-Hall, 1963
Leliavsky, Serge. Arches and Short Span Bridges. New York, New York:
Chapman and Hall, 1982
Leontovich, Valerian. Frames and Arches. New York, New York: McGraw-Hill
Book Company, 1959
Melbourne, C. Arch Bridges. London: Thomas Telford, 1995
Melan, J. Plain and Reinforced Concrete Arches. New York, New York: John
Wiley and Sons, 1915
Nettleton, Douglas A. Arch Bridges. Bridge Division, Office of Engineering,
Federal Highway Administration. Washington D.C. : 1977
Spofford, Charles. The Theory of Continuous Structures and Arches. New York,
New York: McGraw-Hill Book Company, 1937

Page 50 ot 50
MASSACHUSETTS OF TECHNOLOGY
INSTITUTE OF
MASSACHUSETTS INSTITUTE TECHNOLOGY Page 50 of 50