THESIS

STRUCTURAL HEALTH MONITORING MEMS SENSORS

USING ELASTICITY-BASED BEAM VIBRATIONS

Submitted by

Alivia Plankis

Department of Civil and Environmental Engineering

In partial fulfillment of the requirements

For the Degree of Master of Science

Colorado State University

Fort Collins, Colorado

Summer 2012

Master’s Committee:

Advisor: Paul Heyliger

Rebecca Atadero
Robert Leisure
 ABSTRACT

STRUCTURAL HEALTH MONITORING MEMS SENSORS

USING ELASTICITY-BASED BEAM VIBRATIONS

The worsening problem of aging and deficient infrastructure in this nation and across the world has

demonstrated the need for an improved system to monitor and maintain these structures. The field of

structural health monitoring has grown in recent years to address this issue. The goal of this field is to

continually monitor the condition of a structure to detect and mitigate damage that may occur. Many

structural health monitoring methods have been developed and most of these require sensor systems to

collect the necessary information to assess the current strength and integrity of a structure. The motivation

for this thesis is a proposed new microelectromechanical systems (MEMS) sensor with applications in civil

infrastructure sensing. The work required was to determine accurate estimates of the resonant frequencies

for a fixed-fixed silicon bridge within the device so that further testing and development could proceed.

Additional knowledge and information were essential, though, before these requested calculations could be

performed confidently. First, a thorough review of current structural health monitoring concepts and methods

was performed to better understand the field in which this device would be applied and what incentive existed

to develop a new sensor. Second, an in-depth investigation of vibrational beam mechanics theories was

completed to ensure the accuracy of the frequency results for the new MEMS sensor. This study analyzed

the influence of three assumptions employed in the Euler-Bernoulli, Rayleigh, and Timoshenko beam theories

by comparing their results to a three-dimensional, elasticity-based approximation for vibrational frequencies

and mode shapes. The results of this study showed that all three theories are insufficient when a fixed

support is involved, so the elasticity-based approximation was utilized to calculate the frequencies for the

bridge component in the MEMS device. These results have been passed on to the developers so that the

testing process could move forward in the hopes that the device could advance the field of structural health

monitoring in the future.

ii
 ACKNOWLEDGEMENTS

The funding for this research was provided by a grant from Mountain Plains Consortium (MPC). Dr. Paul

Heyliger provided continual support and guidance as the advisor for this thesis work. Dr. Rebecca Atadero

and Dr. Robert Leisure also gave their time and helpful direction as members of the thesis committee. Dr.

Ward Johnson provided the technical information for the proposed MEMS sensor and provided significant

guidance in understanding the operation and application of the device.

iii
 TABLE OF CONTENTS

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii

Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Thesis Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Review of Structural Health Monitoring Concepts and Practices . . . . . . . . . . . . . 5

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.2 Global Health Monitoring Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.2.1 Resonant Frequencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.2.2 Mode Shape Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.2.3 Mode Shape Curvatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.2.4 Dynamic Flexibility Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.2.5 Updating Modal Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.2.6 Acoustic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2.3 Local Health Monitoring Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.3.1 Visual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.3.2 Strains or Displacements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.3.3 Stresses or Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.3.4 Acoustic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.4 Sensors Used in Structural Health Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.4.1 Accelerometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

iv
 2.4.2 Strain Gages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.4.3 Displacement Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.4.4 Tiltmeters and Inclinometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2.4.5 Acoustic Emission Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.4.6 Fiber Optic Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2.4.7 Microelectromechanical Systems (MEMS) Sensors . . . . . . . . . . . . . . . . . . . . 25

2.5 Summarizing Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3 Theory of Beam Vibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.2 Basic Beam Theories in Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

3.2.1 Euler-Bernoulli Beam Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

3.2.2 Rayleigh Beam Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3.2.3 Timoshenko Beam Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3.2.4 Elasticity Based Beam Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.3 Governing Beam Theory Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.3.1 Euler Bernoulli Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.3.2 Rayleigh Beam Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.3.3 Timoshenko Beam Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.3.4 Linear Elasticity Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

3.3.5 Fixed-Free Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3.3.6 Fixed-Fixed Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

3.3.7 Simply-Supported Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4 Methods and Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

4.2 Vibration Study Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.3 Vibration Study Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.3.1 Effects of Slenderness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

4.3.2 Effects of Anisotropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

4.3.3 Effects of Poisson Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

v
 4.3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

4.4 Method for Investigating MEMS Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

4.5 Results from MEMS Sensor Investigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

5.1 Summary of Work and Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

5.2 Suggestions for Continued Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

vi
 LIST OF TABLES

3.1 Euler-Bernoulli Wave Numbers for Fixed-Free Beam . . . . . . . . . . . . . . . . . . . . . . . 46

3.2 Euler-Bernoulli Wave Numbers for Fixed-Fixed Beam . . . . . . . . . . . . . . . . . . . . . . 49

3.3 Euler-Bernoulli Wave Numbers for Simply-Supported Beam . . . . . . . . . . . . . . . . . . . 51

4.1 Material Properties for Vibration Study, units are GPa except for dimensionless νij and Den-

sity in kg/m3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

4.2 Frequencies of an Isotropic Cantilevered Beam . . . . . . . . . . . . . . . . . . . . . . . . . . 57

4.3 Frequencies of an Orthotropic Cantilevered Beam . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.4 Frequencies of an Isotropic Fixed-Fixed Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

4.5 Frequencies of an Orthotropic Fixed-Fixed Beam . . . . . . . . . . . . . . . . . . . . . . . . . 60

4.6 Frequencies of an Isotropic Simply-Supported Beam . . . . . . . . . . . . . . . . . . . . . . . 61

4.7 Frequencies of an Orthotropic Simply-Supported Beam . . . . . . . . . . . . . . . . . . . . . . 62

4.8 Material Properties for Silicon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

4.9 First Five Lateral-Flexural Mode Frequencies for MEMS Bridge . . . . . . . . . . . . . . . . . 80

vii
 LIST OF FIGURES

2.1 Basic Configuration of a Piezoelectric Accelerometer [14] . . . . . . . . . . . . . . . . . . . . . 14

2.2 Structure Types for Piezoelectric Accelerometers [15] . . . . . . . . . . . . . . . . . . . . . . . 15

2.3 Diagram of Servo Force Balance Accelerometer Operation Principle [11] . . . . . . . . . . . . 17

2.4 Basic Configuration of an Electrical Resistance Strain Gage [20] . . . . . . . . . . . . . . . . . 18

2.5 Schematic Diagram of the Operation of a Vibrating Wire Strain Gage [11] . . . . . . . . . . . 19

2.6 External Application Vibrating Wire Strain Gage Model [26] . . . . . . . . . . . . . . . . . . 19

2.7 Schematic of Angular Potentiometer Operation [11] . . . . . . . . . . . . . . . . . . . . . . . . 21

2.8 Linear Variable Differential Transformer Internal Composition Diagram [11] . . . . . . . . . . 21

2.9 Diagram of the Operation of an Inertial Based Inclinometer [30] . . . . . . . . . . . . . . . . . 23

2.10 Diagram of an Acoustic Emission Sensor [31] . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

4.1 Frequency Error for the First Mode of a Cantilevered Beam . . . . . . . . . . . . . . . . . . . 64

4.2 Frequency Error for the Fifth Mode of a Cantilevered Beam . . . . . . . . . . . . . . . . . . . 64

4.3 Frequency Error for the First Mode of a Fixed-Fixed Beam . . . . . . . . . . . . . . . . . . . 65

4.4 Frequency Error for the Fifth Mode of a Fixed-Fixed Beam . . . . . . . . . . . . . . . . . . . 65

4.5 Frequency Error for the First Mode of a Simply-Supported Beam . . . . . . . . . . . . . . . . 66

4.6 Frequency Error for the Fifth Mode of a Simply-Supported Beam . . . . . . . . . . . . . . . . 67

4.7 First Five Mode Shapes for a 1x1x40 Isotropic Fixed-Free Beam . . . . . . . . . . . . . . . . 69

4.8 First Five Mode Shapes for a 1x1x2 Isotropic Fixed-Free Beam . . . . . . . . . . . . . . . . . 69

4.9 First Five Mode Shapes for a 1x1x40 Isotropic Fixed-Fixed Beam . . . . . . . . . . . . . . . . 70

4.10 First Five Mode Shapes for a 1x1x2 Isotropic Fixed-Fixed Beam . . . . . . . . . . . . . . . . 70

4.11 First Five Mode Shapes for a 1x1x40 Isotropic Simply-Supported Beam . . . . . . . . . . . . 71

4.12 First Five Mode Shapes for a 1x1x2 Isotropic Simply-Supported Beam . . . . . . . . . . . . . 71

4.13 Error Comparison Between Isotropic and Orthotropic Materials for a Cantilevered Beam . . . 72

4.14 Error Comparison Between Isotropic and Orthotropic Materials for a Fixed-Fixed Beam . . . 72

4.15 Error Comparison Between Isotropic and Orthotropic Materials for a Simply-Supported Beam 73

4.16 Influence of Poisson Effect for Varying Slenderness Ratios . . . . . . . . . . . . . . . . . . . . 75

4.17 Influence of Poisson Effect for Varying Modes of Vibration . . . . . . . . . . . . . . . . . . . . 76

viii
4.18 Illustration of MEMS Sensor Configuration Cross-Sectioned Through the Mid-Length . . . . 78

4.19 Illustration of Cross Section Deformation for Through-Thickness Shear Mode . . . . . . . . . 80

4.20 Lateral-Flexural Vibration Modes for MEMS Bridge . . . . . . . . . . . . . . . . . . . . . . . 80

ix
Chapter 1

Introduction

The issue of aging infrastructure in this country has become apparent to the engineering community in

recent years due to catastrophic failures, such as the collapse of the I-35W bridge across the Mississippi River

in Minneapolis, MN and also from an increasing awareness of the current state of structural health. Recent

structural surveys have shown more than a quarter of the nation’s bridges to be structurally deficient or

obsolete and have brought to attention the fact that more than a third have been in service for more than

40 years [1, 2]. This data, along with the occurrence of major failures, has demonstrated the necessity for

methods in which to detect structural damage and properly prioritize the replacement of existing bridges and

buildings. A notable field which has developed to answer this need is that of structural health monitoring.

The basic concept behind this field is that by continually tracking the strength and integrity of a structure,

damage can be detected, the life span can be assessed, and proper maintenance and mitigation actions taken.

In order to perform most structural health monitoring methods, sensors are required to collect data on the

key parameters involved in evaluating a structure’s state of strength. Numerous devices have been developed

and applied for this purpose, but a great deal of advancements still need to be made before structural health

monitoring methods can be applied prevalently in an accurate and cost efficient manner.

The motivation behind this thesis is a proposed new microelectromechanical systems (MEMS) sensor with

applications in monitoring civil infrastructure, which could help address some of the deficiencies with current

sensor technologies. A thin fixed-fixed silicon bridge provides the primary sensing element in this device.

As the sensor is loaded, the resonant frequency of the bridge will change and provide information about the

component to which it is attached. The key goal of this thesis was to investigate the properties of this bridge

by performing calculations for certain modal frequencies which are needed before further development can

proceed. In advance to successfully executing the requested analysis, additional understanding was essential.

To better comprehend the incentive for the invention and development of a new sensor, a review of current

1
structural health monitoring methods and the technologies utilized for them was conducted. Also, to ensure

more accurate frequency results an in-depth vibrational mechanics study was completed to determine the

effectiveness of common beam theories for varying support conditions. The remainder of this chapter will

layout the objectives of each of these sections, how they were obtained, and the organization of the paper.

1.1 Objectives

This thesis contains three main goals which are all connected, but each have their own objectives. The

goal which was the impetus for this research was to determine accurate modal frequencies for the silicon

bridge within the proposed MEMS sensor. The details and methods for this goal will be discussed last

because the completion of the other two sections of this thesis was necessary before this goal could be

achieved.

The first basic goal for this thesis was to obtain a greater understanding of the field of structural health

monitoring through a review of concepts, methods, and sensors. With the knowledge gained from this

research, the motivation for a new sensor could be recognized and thus provide defense for the relevance of

the calculations for the new MEMS sensor. A number of objectives were created to ensure that this study

of structural health monitoring was successful and that adequate information was gained. These objectives

are listed below.

1. Ascertain the primary goals of the practice of structural health monitoring.

2. Identify common methods which have been developed to attain these goals.

3. Explain the concepts behind each of these methods, as well as both the benefits and drawbacks asso-

ciated with their use.

4. Establish what types of sensors are available or in development to collect the necessary data for these

structural health monitoring methods to be effective.

5. Describe the theory of operation for these sensors, how they are applicable, and any advantages or

disadvantages they may possess.

The objectives provided were achieved by a process of thorough research of reference books, journal

articles, academic reports, and more. The topic of this research was at first very broad, covering the basics

of structural health monitoring, and was continually refined as more detail of methods and sensors became

2
of interest. The amount of information available is immense, so the review created is not comprehensive, but

the material found and discussed provided a sufficient knowledge base to move forward with the calculations

for the suggested new sensor, and possibly provide aid for others who may continue the development and

research for this and other similar devices.

The second leading goal of this thesis was to perform a vibrational mechanics study, the results of which

could be used to ensure the accuracy of the frequency calculations for the MEMS device. This vibration

study considered three commonly used beam mechanics theories: Euler-Bernoulli, Rayleigh, and Timo-

shenko. These three models increase in complexity, but all contain common assumptions to simplify analysis

procedures. The purpose of this investigation was to determine how three of these common assumptions

affect the accuracy of frequency and mode shape calculations. In order to achieve this, a more accurate

three-dimensional elasticity approximation was utilized. The following objectives were conceived to carry

out this study effectively.

1. Present the theory behind each of the common beam models and the displacement effects which are

included or ignored in each.

2. Provide a list of the common assumptions included in all three of the discussed beam theories.

3. Explain the three-dimensional elasticity approximation concepts and its development.

4. Investigate the effects of three of the basic assumptions: high slenderness, isotropic material properties,

and the idea that the Poisson effect is negligible.

5. Analyze and discuss if and when each of the three common beam theories are applicable based upon

the calculations.

A number of steps were taken to accomplish these objectives. First, computational programs, which

had been provided from previous research, were modified to produce frequency calculations for each of the

beam theories and the elasticity solution when applied to three different support cases: fixed-free, fixed-

fixed, and simply-supported. Once these were prepared various calculations were performed to allow for the

investigation of the three assumptions of interest. Varying beam lengths were employed to study slenderness

and two sets of material properties were applied to research the effect of anisotropy. Lastly, the more accurate

elasticity based solution was utilized for calculations both with and without the Poisson ratio included to

better isolate this effect. The conclusions determined about the accuracy of common beam theories from

3
this vibrational mechanics study were applied to the calculations performed on the fixed-fixed bridge of the

MEMS sensor.

The final goal of determining accurate vibrational frequency estimates for the new MEMS sensor was

possible after the knowledge gained from the previous two goals was applied. The more accurate three-

dimensional elasticity based approximation method was employed for the frequency calculations. Accurate

results were obtained for varying bridge lengths and passed on to the developers of the device. The main

objective of this section was to provide valuable information to allow for further development and research

to proceed for this new and promising device.

1.2 Thesis Layout

The information for the three goals of this thesis has been organized into the four remaining chapters.

Chapter 2 presents the results of the structural health monitoring review. It covers two categories of possible

methods, with specific examples for each of them which are explained and compared. The types of sensors

which are available are then explained and discussed. The information provides a thorough understanding

of the field of structural health monitoring, how it is currently performed, and what improvements can be

made.

In Chapter 3, all of the concepts and development of the three beam theories to be studied and the

proposed full elasticity based solution are explained. The mechanics effects which are included in each of

them are presented and the governing equations that come out of these inclusions are provided and discussed.

Each of the models are compared and contrasted to understand the limitations and advantages of each of

them. This chapter also covers the changes necessary for each of the theories to be applicable to the varying

support conditions which were analyzed.

Chapter 4 discusses the details of the methods employed for the vibrational mechanics study and the

MEMS sensor calculations. The results of the study are presented and examined in detail, with final con-

clusions produced and listed. The final frequency calculations for the new MEMS device are also displayed

in Chapter 4.

The final chapter of this thesis provides concluding remarks. Summarizing statements regarding each of

the three main goals of this thesis are produced and suggestions for further research which could stem from

the work in this thesis are proposed.

4
Chapter 2

Review of Structural Health

Monitoring Concepts and Practices

2.1 Introduction

The bridges and buildings that society uses and depends upon daily do not retain their strength and

utility indefinitely. These structures deteriorate over time as they are continually exposed to harsh environ-

mental conditions and continued or increasing loads. The issue of aging infrastructure has gained increasing

interest over the last decade due to a few catastrophic failures as well as an expanding awareness of current

infrastructure conditions. In August of 2007, the Interstate 35W Bridge across the Mississippi River in Min-

neapolis, MN suddenly collapsed, killing 13 people and injuring 144 more [1]. This bridge was 40 years old

and had been subjected to increased traffic and environmental loads over its lifetime, causing deterioration

and ultimately the failure of some under-designed components. The I-35W Bridge is not an entirely unique

case either. Much of the country’s civil infrastructure is significantly aged and in need of attention. More

than 35% of U.S. bridges were built more than 40 years ago [1] and according to the U.S. Department of

Transportation (USDOT), in 2009, nearly a quarter of the country’s bridges were either structurally deficient

or structurally obsolete [2]. This growing problem needs to be addressed and one of the ways engineers are

doing so is through structural health monitoring. This is the process of tracking the condition of a structure

over time to determine the nature and extent of damage that may exist and how this affects the integrity of

the structure. Knowing this information can greatly improve the maintenance process and possibly prevent

future damage or even collapse of civil infrastructure.

The field of structural health monitoring has grown significantly in recent years. There has been sub-

stantial research on the subject and there are continually more projects beginning in this field. The primary

5
goal of structural health monitoring is to identify damage in a structure. Within damage identification, four

different levels exist. The first level determines that damage is present within the structure. The second level

includes locating the damage, while the third level assesses the extent of this damage. The final level uses the

information from the first three steps to predict the remaining service life of the structure [3]. The following

review will focus on methods which deal with the first three steps. There are two main method categories

which exist in structural health monitoring to address these levels of damage identification. First, there

are methods that are used for global monitoring. These methods result in information about the structure

as a whole and how it responds to loads. The second category is local health monitoring methods. These

methods are used to monitor local behavior at critical points on the structure. With local methods, the areas

where damage is most likely must be known ahead of time for the analysis to be relevant. Depending on

the information an engineer is interested in about a structure, and under which of these two categories that

falls, there are multiple systems and types of sensors that can be employed to collect the desired information.

This chapter will review the basic concepts behind both categories of structural health monitoring and the

systems that have been considered or utilized to perform the analysis.

2.2 Global Health Monitoring Methods

The basic concept in global health monitoring methods is that the existence of damage will change the

stiffness, mass, or damping properties of a structure and thus alter the global dynamic properties. The major

benefit of this type of method is that the location of possible damage does not need to be known before

installing a system, which is often not possible for complex structures. Methods of global health monitoring

can be based on a number of different modal properties of a structure. The benefits and challenges to

techniques based on each of these are discussed below. Also discussed are some methods based on acoustic

properties of materials rather than modal properties of the structure.

2.2.1 Resonant Frequencies

Modal frequencies are one of the most basic vibrational properties of a structure. Every structure

possesses unique resonant frequencies for each mode of vibration that are related to its mass and stiffness.

This basic property can be used to detect the occurrence of damage by noting shifts in resonant frequency

values. Although this is a very appealing concept, there are quite a few limitations that restrict the practical

application of this method.

6
 Resonant frequencies are a global parameter and do not provide any spatial information about a structure.

It is evident that methods based on frequency shifts could only be effective at detecting the existence of

defects and would require the use of an additional analysis technique to locate and assess the severity of

any damage that may occur. Also, the sensitivity of frequency to damage is relatively low, especially for

large structures. Significant damage must occur in the structure to cause a detectable change in resonant

frequency for most measurement systems. A very precise monitoring system could possibly detect shifts

from minor damage, but it then becomes difficult to separate shifts due to damage from changes caused

by environmental and operational factors. Significant efforts were put forth to develop a damage detection

method based on frequency shifts for offshore oil platforms during the 1970s and 1980s, but these were

abandoned after encountering issues with machine noise, temperature variations, and changing mass due to

varying fluid levels and marine growth [4]. It was impossible to separate changes caused by damage from

all of these other factors. Resonant frequencies at higher modes where the modes are associated with local

responses become more sensitive to minor damage, but it is often unrealistic to excite and determine these

modes. The one benefit of frequency based methods over other global methods is that resonant frequencies

could be found with less uncertainty than other modal parameters [5].

Many frequency based health monitoring methods assume that cracking is the only form of damage that

will occur. This significantly limits application possibilities. In concrete structures, for example, most of

the stiffness is provided by the concrete and thus damage or deterioration of the reinforcing steel would not

have a significant effect on the natural frequencies of the structure [6]. Using a frequency based method

may be able to detect cracks in the concrete, but would miss reinforcement and confinement damage issues.

Similarly, in steel structures even a significant amount of corrosion may not have a detectable effect on the

stiffness and would not cause a noticeable change in frequency. Therefore, frequency based health monitoring

methods can be effective in detecting significant damage from cracking, but they are not comprehensive and

other methods are needed to adequately monitor the health of a structure.

2.2.2 Mode Shape Vectors

The mode shape vectors associated with each mode of vibration are another basic dynamic property of

a structure. These are also related to the mass, stiffness, and damping of a system and are affected when

damage changes any of these parameters. A major advantage of damage detection methods based on mode

shape over those based on frequency is the ability to locate the damage that occurs. Mode shape vectors are a

spatial property and can provide information about both the existence and location of damage. The difficulty

7
with this method is that it requires a significant number of measurement locations to accurately determine

the mode shape and any changes caused by damage. Similar to frequency based methods, modal shape

monitoring methods also suffer from low sensitivity issues. Most damage that occurs is a local phenomenon

and cannot be easily detected by studying the lower frequency mode shapes that are typically observed in

large structures. In fact, mode shape changes can be even more subtle than frequency changes in typical

structural response [6]. Mode shapes of higher frequency modes would be more sensitive to minor local

damage, but as discussed before these modes are difficult to induce. Mode shape based methods possess

most of the same limitations as frequency based methods. A significant amount of damage must occur to

be detectable and many types of damage would not have a relevant effect on mode shapes. Overall, mode

shape methods have the added benefit of the ability to locate damage, but are still insufficient for minor

levels of damage and deficiencies other than cracks.

2.2.3 Mode Shape Curvatures

An alternative to using mode shapes to detect damage is to instead consider the mode shape curvature.

The curvature is found by differentiating the mode shape vector twice. The value of curvature at a point

in structure is equal to M/EI, so if the stiffness at a point is reduced by damage, the curvature at that

point will increase. This can be used to both detect and locate damage in a structure as well as possibly

estimating the extent of damage by studying the amount of fluctuation in curvature values [7]. Studying the

curvature is an improvement over using mode shapes due to the fact that the curvature is far more sensitive

to small changes in the structure than the mode shape. Thus, minor levels of damage may be able to be

detected even in lower frequency modes with curvature based methods. Another advantage is that there

exists a direct relationship between mode shape curvature and bending strain for beams, plates, and shells.

Bending strain has also been shown to be far more sensitive to low levels of damage than frequency and mode

shape based methods [4]. The greatest drawback of curvature methods is the accuracy. The techniques used

for finding the second derivatives of the mode shapes can cause false readings of damage when none exist,

which contaminates the results [8]. Also, it was found that the statistical uncertainty in finding curvature

values was greater than in determining mode shape vectors or resonant frequencies [4]. Another difficulty

with curvature based monitoring methods can occur if there is not baseline data available for a structure. If

damage is distributed throughout the structure before any data is collected it may be challenging to detect

damaged locations without knowing what the curvature values were before damage occurred [6]. Mode shape

curvature methods are a more realistic option for large civil infrastructure health monitoring in that it only

8
requires information from low frequency modes to be effective, but there are still accuracy and applicability

issues that need to be resolved.

2.2.4 Dynamic Flexibility Matrix

Another method of global monitoring methods is based on differences in the dynamically measured

flexibility matrix. The flexibility matrix is the inverse of the stiffness matrix for a structure and thus relates

applied static force to structural displacement. The flexibility matrix is calculated from the mass normalized

mode shapes and frequencies of a structure. Each column of this matrix represents the displacement pattern

for a certain degree of freedom caused by a unit action applied to that degree of freedom. By studying the

changes in the dynamic flexibility matrix over time, the existence and location of damage can be detected.

The degree of freedom that displays the maximum variation in flexibility from the undamaged state represents

the location of possible damage [7]. One of the main advantages of this method is that the flexibility matrix

is inversely related to the modal frequencies, which makes it most sensitive to the lower-frequency modes

that are more dominant in large structures [3].

There are some drawbacks to using a flexibility matrix method. In most applications only a few of the

mode shapes will be calculated, which means that the flexibility matrix will only be approximate since its

proper calculation requires all of the possible mode shapes. Also, this method works best when looking at

flexibility changes between the completely undamaged state and the current condition. This requires either

that measurements be taken immediately after construction of a structure or the use of a finite element

model of the original structure. In many cases, researchers are interested in studying an already existing

structure, so initial and undamaged measurements are impossible. Using a model of the original structure

is a possible solution to this issue, but multiple assumptions usually need to be made to build the model

which makes the analytical undamaged flexibility values approximate. Damage detection methods based

on the dynamic flexibility matrix of a structure are an improvement over frequency and mode shape based

methods since they are more sensitive to lower frequencies, but may have accuracy issues due to the amount

of approximation involved in the analysis.

2.2.5 Updating Modal Parameters

Another method to detect damage in structures involves updating modal property matrices such as

mass, stiffness, or damping of a structural model to match the measured responses as closely as possible.

These methods estimate the updated modal matrices through constrained optimization techniques based

9
on equations of motion, the structural model, and the measured data [3]. These updated matrices are

then compared to the original undamaged values to detect and locate damage. These methods are not as

common as the previously discussed options because a number of problems exist with them. The original

matrices for the undamaged case may be inaccurate to begin with and the optimization process does not

produce a unique solution, so the updated matrices may not be representative of what is actually occurring

in the structure. Also, the error minimization operation may cause the stiffness of undamaged elements to

change which could lead to an abundance of false damage results [6]. Often the degrees of freedom from

the analytical model matrices do not match the locations of measurement which can severely reduce the

applicability or effectiveness of these methods [4]. Updating the modal properties of a structure is another

option to detect damage that may have occurred, but there are significant obstacles to overcome with these

methods.

2.2.6 Acoustic Properties

Another class of structural health monitoring methods is based on acoustic properties rather than modal

properties. These methods are based on the measurement of waves propagating through the structural

material. Some methods are based on how sound waves travel through different material anomalies, but

these are mostly only employed in local health monitoring and will be discussed in the next section. Acoustic

monitoring techniques, such as acoustic emission systems, can also measure how stress waves radiate through

a structure. The basic concept behind acoustic emission methods is that stress waves are produced by a

sudden release of energy due to micro-cracking occurring in the material. These stress waves can be picked

up by acoustic emission sensors to detect cracking and even locate it through triangulation with multiple

sensors [9]. A significant advantage to this type of global health monitoring is that it has the ability to detect

damage events as they occur rather than after some analysis of the data. This can present a complication as

well. This method is only applicable if data is being collected continuously, otherwise the stress emissions

may be missed entirely. Continuous data collection may not be possible or cost effective in many cases. Some

other drawbacks of this method are that the waves emitted by the micro-cracking can be very weak and

difficult to detect over background noise and determining what type of damage has occurred is challenging

[10]. Acoustic emission monitoring methods are a popular option because they can theoretically detect

damage anywhere in a structure as it is happening. This could have significant applications in early warning

systems that aim to shut down or evacuate structures if damage is indicating possible catastrophic failure.

10
2.3 Local Health Monitoring Methods

Once potential damage locations are determined from global monitoring methods, or if critical areas are

initially known, local health monitoring methods are performed to assess the extent of damage and keep

track of any damage progression. There are numerous techniques for performing local damage detection and

assessment, but they all center on a few basic concepts. The properties measured and concepts employed in

local health monitoring methods are discussed in this section.

2.3.1 Visual

Probably the most common type of local health monitoring practiced is visual inspection. This extremely

simplistic method identifies damage in a structure simply by looking at it and hopefully recognizing changes

and abnormalities. These methods are used exceedingly often primarily due to the biannual bridge inspec-

tions mandated by the National Bridge Inspection Program enacted in 1967 [11]. This program requires

current information on the nearly 600,000 bridges across the United States and the adopted procedure for

completing this requirement is visual inspection. These traditional techniques require immense amounts of

time, closure of a bridge in many cases, and a great deal of manpower and other costs. Visual inspections

also depend highly on the decisions and opinions of the inspector and can lead to large variations and sub-

jectivity in the results. Furthermore, only outward appearance of a structure can be assessed, so even major

internal damage could be missed for years [10]. Therefore, although this is still a popularly used method, it

has multiple flaws and more objective and comprehensive methods are needed to safely assess the current

health of structures.

2.3.2 Strains or Displacements

Monitoring and assessing damage through either strains or displacements provide another favored option.

As a structure is loaded and stressed the material goes through strains and deflections in response, and if

these strains or deflections exceed certain values damage can be assumed. This type of method is typically

used for predicted critical locations of a structure such as a connection or a potential weak spot that has

been detected by a global detection method. Another monitoring application utilizing strain or displacement

measurements is to watch an already known defect such as a crack to determine if the damage is worsening.

Another parameter similar to displacement that may be monitored is the tilt or slope of a member. This

is typically used to study pier behavior [10]. As environmental or loading conditions change the angle of

11
a bridge support may change which could indicate weakness or failure. The primary problem with these

methods is how localized their measurement capabilities are. Often strain is only measured over a few

inches or less and displacement and tilt are only recorded at one point on the structure. In general, though,

monitoring methods based on strains or displacements are effective and commonly used.

2.3.3 Stresses or Loads

Another concept used in local health monitoring measures the stress or load which a part of the structure

experiences. In relation to local health monitoring, if the stress or load occurring in a certain location

is beyond a certain level damage may be likely. Conversely, if the stress or load measurement decreases

significantly or rapidly, it may be a signal that damage has occurred and the structure can no longer withstand

the load it was previously supporting. These methods are utilized much less often because stresses or loads

are substantially more difficult to directly measure than strain or displacement.

2.3.4 Acoustic Properties

Local health monitoring methods may also use acoustic properties to detect or measure damage. A few

methods consider the propagation of sound waves through the material. The reflection of sound waves will

change if anomalies such as air voids exist in the material. This principle is the basis for both chain dragging

and tap tests. Chain dragging is used to determine bridge deck health and is a widely used and accepted local

monitoring method. Tap tests are typically used to detect delamination between fiber reinforced polymer

(FRP) sheets and concrete. Both of these methods possess subjectivity problems since the results depend on

the inspector that is listening to the response. Two different inspectors can find entirely different conclusions.

Another disadvantage with these two techniques comes from access concerns. With chain dragging, complete

bridge closure is often required [10] and reaching testing locations for tap tests may be unsafe or impossible.

Another acoustic based method employs the concept that stress wave propagation from a force is affected

directly by mechanical properties [9]. This type of method, referred to as impact echo, involves applying a

minor impact force to a structure and listening to its response. Based on the frequency of the wave response

delaminations, voids, and cracks can be detected as well as the depth at which they exist. This method has

been found to be highly accurate, which is a significant advantage. A main disadvantage to this method is

that many locations need to be tested in order to obtain a comprehensive understanding of the defects that

may exist. Also, when applied to bridges this technique requires that no traffic is present during testing, so

lane closures are necessary [10].

12
2.4 Sensors Used in Structural Health Monitoring

All of the health monitoring methods discussed require data values from the structure of interest to

perform analysis and determine if any damage has occurred or if damage is probable. Although a few of

the methods presented only require data from visual or manual inspection, this type of data is tedious

and costly to obtain. Because of this issue, most research in the area of structural health monitoring has

focused on the use of sensor systems to collect the necessary information for analysis. Two main types of

quantities are typically measured with structural sensors: kinematic and environmental [12]. Environmental

metrics such as temperature and humidity are commonly monitored to isolate the response of structure due

to loading from these effects. Usually, the values of primary interest are of the kinematic type such as strain,

displacement, and force. Therefore, the remainder of this section will focus on sensors that measure only

kinematic quantities while realizing that these sensors would need to be combined with environmental data

to properly analyze the state of a structure.

A multitude of different sensors have been developed and discussed for the use of structural health mon-

itoring, the descriptions of which would fill an entire book at least. The following sections will present

the most commonly used sensors which have been well established and tested as well as some very promis-

ing recently developed sensor technologies, but this list is by no means a comprehensive presentation of

available sensor technologies. Sensors which measure the dynamic vibrational properties required for global

health monitoring will be presented first, then local kinematic sensors, and finally concluding with recent

developments in the field of health monitoring sensors.

2.4.1 Accelerometers

Accelerometers measure the acceleration a particular location of a structure is experiencing due to either

gravity or applied loads. Different versions of accelerometers have been used as measuring devices for

many years because acceleration data can provide valuable information about the dynamic characteristics

of a structure. Through post processing procedures, acceleration measurements can be used to calculate

frequency, damping, and mode shapes of a structure [11]. As discussed previously, these vibrational properties

are useful in global health monitoring methods. Although accelerometers have long been employed for

monitoring techniques, issues still exist with error development during the numerical integration of the data

[13]. These devices also output a great deal of data which can require intensive processing. Even with these

challenges, accelerometers provide the best option for measuring dynamic properties of a structure. Four

13
different types of accelerometers are commonly used and each have their own advantages and drawbacks

which will be discussed in the following paragraphs.

Piezoelectric Accelerometers

The most popular type of accelerometer of the four common options is the piezoelectric accelerometer.

These sensors have many advantages including small size, high output, durability, wide frequency range,

and the ability to monitor a wide range of dynamic events [14, 11]. The basic components of a piezoelectric

accelerometer are the base, a piezoelectric element, and a seismic mass. A basic set up for this type of

sensor is shown in Figure 2.1. The piezoelectric elements in these sensors are the key components in the

operation of the sensor. Piezoelectric materials have the unique property that they output an electrical signal

proportional to the stress applied to the material [14]. As an acceleration is applied to the base of the sensor,

the piezoelectric elements experience a force proportional to the mass they are connected to and thus to the

acceleration being experienced. This force creates an electrical output that can be sent through a cable to a

data acquisition system where it can be processed further. A significant advantage to these types of sensors

is that they are self-generating. The piezoelectric material outputs an electrical signal without the need of

a constant input electrical or power source [14]. When this type of accelerometer was originally invented, it

required special low-noise cabling because the original signal from the piezoelectric material is susceptible to

corruption from environmental effects and cable-generated noise [14]. Today, these accelerometers usually

include electronics within the sensor which covert the signal so that it can be transmitted over long cable

distances without any loss in quality [11]. The only downside to this addition is that it somewhat limits the

temperature range in which the sensors can be used, but they still operate up to 350◦ F [14].

Figure 2.1: Basic Configuration of a Piezoelectric Accelerometer [14]

14
 There are a few different options available with piezoelectric accelerometers. Two types of piezoelectric

materials exist which can be utilized in these sensors: quartz or polycrystalline ceramics. Quartz is a natural

material that is inherently piezoelectric and is therefore considered more stable than other material options

and provides a better option for long-term accelerometers. Quartz is also immune to the the pyroelectric

effect, so the output is not affected by temperature change. One drawback to quartz is that it has low

charge sensitivity which limits its use in certain systems. Polycrystalline ceramics are man-made and forced

to become piezoelectric through the process of poling. If these materials are exposed to drastic temperature

changes or large electric fields their properties can be significantly altered which is their biggest disadvantage.

Polycrystalline ceramics provide the benefit of being easily adapted to different properties, unlike quartz.

The ceramics can be created to have high charge sensitivity or to withstand extremely high temperatures [14].

Both types of material have their usefulness, it simply depends on the application the sensor it to be used

for. Three different structures of piezoelectric accelerometers are also available. The structure options are

based upon the type of stress applied to the piezoelectric material in the sensor. The options are shear mode,

flexural mode, and compression mode. Diagrams of each of these structure types are shown in Figure 2.2.

The shear mode assembly allows for very small size, which is useful for high-frequency response. Flexural

mode accelerometers are best for low-frequency and low-gravitational accelerations and are not sensitive to

transverse motion. The last option, compression mode, is very rugged and can withstand high shock levels

but can be more sensitive to thermal and strain effects since the piezoelectric material is in contact with the

base [14].

(a) Shear Mode (b) Flexural Mode (c) Compression Mode

Figure 2.2: Structure Types for Piezoelectric Accelerometers [15]

15
Piezoresistive Accelerometers

Another common accelerometer type utilizes piezoresistive strain gages. These strain gages use resistors

made of single-crystal silicon. The electrical resistance of this material changes in proportion to the stress

or force applied to it [11]. The common structure of piezoresistive accelerometers involves these strain gages

attached to a cantilever bridge with a seismic mass attached. As the sensor experiences acceleration, the

strain gages are stressed in flexure which alter the electrical signal output by a Wheatstone bridge circuit. The

main advantage to this type of accelerometer is that they can be produced in a very small size, which makes

them useful in MEMS systems which will be discussed in Section 2.4.7 [14]. Piezoresistive accelerometers also

provide very low frequency response capabilities which is another significant advantage. The main drawback

to this sensor type is that they have a smaller dynamic range of measurement and they require a constant

electrical input for the Wheatstone bridge circuit.

Capacitance Accelerometers

Capacitance accelerometers are very similar to piezoresistive accelerometers, since they measure an elec-

trical change across a bridge, but these sensors measure a change in capacitance rather than resistance [14].

These accelerometers have very high resolution and can measure accelerations from the level of micro-g

up to 100s of g which is better than the resolution available from piezoresistive accelerometers [11]. Some

disadvantages include a limited high-frequency range and a higher noise level than typical piezoelectric

accelerometers.

Servo Force Balance Acclerometers

The last type of acclerometer that can commonly be found is the servo force balance type. The operation

of these accelerometers differs greatly from the other three options discussed. All three previous sensor

types measure the deflection of a seismic mass, which is proportional to acceleration, directly through

different electrical principles. Servo accelerometers instead keep the movement of an internal capacitor

plate to a minimum by using electromagnetic forces to keep the mass in a neutral position. The required

electromagnetic force necessary is proportional to the acceleration experienced [14]. The operation of this

sensor occurs rapidly enough that there is virtually no relative movement between the plate and the supports

which reduces the effects of nonlinearities that can occur with the other three sensor types [11]. This type

of accelerometer is far more accurate than the other three with measurement capabilities on the level of

milli-g. The primary disadvantage to this option is the cost, which can be up to ten times greater than

16
other accelerometers. A figure of the basic principle behind servo force balance accelerometers is provided

in Figure 2.3.

Figure 2.3: Diagram of Servo Force Balance Accelerometer Operation Principle [11]

2.4.2 Strain Gages

Strain gages are commonly used in the practice of structural health monitoring for a number of reasons.

The greatest advantage of these sensors is that they are inexpensive and easy to install [16]. Although strain

gages can only measure the strain of a structure at one location, the information gained from these sensors

can be used in combination with a finite element model of the structure to help detect when damage may

have occurred [17]. The use of strain gages in this type of method are based upon the principle that damage

to critical location will lead to a change in load path and thus a detectable difference in strain distribution

[18]. The use of strain gages in structural health monitoring in this way is common partly due to the fact

that thorough research has been performed involving the interpretation of strain data and this process is

well understood [16]. Many varieties of strain gages exist, but two types are most commonly utilized in

structural health monitoring methods and will presented in the following paragraphs.

Electrical Resistance Strain Gages

Bonded foil resistance strain gages, also known as electrical resistance strain gages, are the most commonly

used type of strain gage [11]. The principle behind these sensors is the fact that the electrical resistance

of a conducting material is proportional to the length and inversely proportional to the cross sectional area

17
[19]. These strain gages consist of a metallic foil, usually arranged in a pattern to increase the length of

the conductor, attached to a nonconducting film which is then connected to the structure of interest with

an epoxy. A diagram of this basic setup is shown in Figure 2.4. With this set up, the strain gage will

deform with the member, and as the foil in the gage is stressed the length and cross sectional area will

change leading to a change in electrical resistance which can be measured [19]. The relationship between

the change in electrical resistance and the change in length of the foil is known as the gage factor, which is

unique to each sensor. With this factor, as well as basic constitutive laws, the relationship between strain

and resistance can be determined. Because electrical resistance is also altered by changing temperatures a

temperature control strain gage is commonly used along with the gage attached to the structure to isolate

the strain due to kinematic loading.

Figure 2.4: Basic Configuration of an Electrical Resistance Strain Gage [20]

Electrical resistance strain gages have a number of advantages and possible applications. These sensors

can be employed on any structure material including steel, concrete, and composites. When applied to

concrete members, the gage length tends to be longer to better average out the localized strain values due

to discontinuities in the material, but this is a simple adjustment. Many other types of sensors such as load

cells, accelerometers, and more also utilize electrical resistance strain gages within their set up, so these

gages can be applied for more than just measuring local strain values [11]. Another advantage to electrical

resistance strain gages is that they have been found to agree well with other types of strain gages in short

term monitoring situations [21, 22]. These sensors can also be used to determine principle strains when

a strain gage rosette is employed [23]. Generally this type of strain gage is only used for short duration

measurements such as the strains caused by traffic loads because they lose their accuracy over time due to

their susceptibility to adverse effects from exposure to the elements. A few other issues exist with electrical

18
strain gages besides their loss of accuracy over time. These sensors are ideally only sensitive to strain in

the longitudinal direction of the gage, but they can be affected by transverse movement as well, which can

reduce the accuracy [11]. They also require a constant power source to be operated and are more susceptible

to noise effects than other types of strain gages. Even with these drawbacks, electrical resistance strain gages

remain the most common option when using strain measurements in structural health monitoring.

Vibrating Wire Strain Gages

Vibrating wire strain gages are the other popular option for strain measurements. These sensors are

known to provide excellent performance and long-term durability and can be installed externally on steel

or concrete or embedded within a concrete member [11]. These sensors include a pretensioned steel wire

which is fixed at both supports and an electromagnetic coil which is used to excite the wire and read its

natural frequency. The wire is encased within a steel tube which can be applied to the desired member

at each end. A diagram of the basic setup for this type of sensor is given in Figure 2.5 and a photograph

of an external application option of this strain gage type is shown in Figure 2.6. This type of strain gage

works by comparing the change in frequency of the pretensioned wire as the length of the wire changes to

determine the strain experienced [24]. These strain gages have been shown to very stable over long periods

of time which makes them applicable for continuous measurement of slowly changing strains unlike electrical

resistance strain gages [25]. The main disadvantages to this sensor option are that they are not applicable

for measuring dynamic strains and there is the possibility for unequal strains to develop in the gage than

in the member due to differences in thermal expansion [22]. Overall, these strain gages provide a reliable

option for long-term strain measurements in structural health monitoring.

Figure 2.5: Schematic Diagram of the Operation of a Figure 2.6: External Application Vibrating Wire
Vibrating Wire Strain Gage [11] Strain Gage Model [26]

19
2.4.3 Displacement Sensors

Displacement is another measurement parameter that is often of interest in the practice of structural

health monitoring. Monitoring of displacements is commonly used to assess critical locations such as across an

expansion joint or over the length of a suspension cable in a bridge [10]. The two most common sensors utilized

to measure displacement are linear potentiometers and linear variable differential transformers (LVDTs). The

details of each of these are presented in the following paragraphs.

Linear Potentiometers

Linear potentiometers, also known as cable extension transducers, are comprised of a spool with a length

of stainless steel cable wound on it, a tension spring that keeps the spool loaded so that either direction

of displacement can be measured, and a precision potentiometer. Usually, the main housing of the sensor

is attached to a fixed location, with the end of the cable connected to the member or component whose

displacement is being monitored. As the location of interest displaces, the cable extends from or retracts

onto the spool which rotates. The rotation of the spool moves the potentiometer. A potentiometer consists

of a thin film resonator with a movable wiper which is in contact with the electrical resistor. As the wiper

moves across the resonator due to the rotation of the cable spool, a change in electrical resistance occurs

which is proportional to the displacement of the cable [11]. A diagram of the basic operation for an angular

potentiometer is shown in Figure 2.7. The benefits of this type of displacement sensor include a low cost,

ability to measure large displacements, simplicity in operation, ease of installation, and higher accuracy and

measurement range than LVDT sensors [11, 27]. Linear potentiometers also have some disadvantages in

comparison to other displacement sensors in that they are not applicable for dynamic measurements or long-

term continuous measurement due to temperature related drift issues, and they are subject to wear problems

over time because of the contact between the potentiometer wiper and resonator [11]. These displacement

sensors still remain a positive option for accurate displacement measurements, especially if the expected

displacements are large.

Linear Variable Differential Transformers (LVDTs)

Linear variable differential transformers (LVDTs) offer another common displacement sensor option.

LVDT sensors are based upon the principle of mutual inductance. These devices consist of a hollow metallic

cylinder containing a primary and two secondary electrical coils and a movable magnetic core. A diagram

of this set up is provided in Figure 2.8. For the operation of this sensor a voltage is applied to the primary

20
 Figure 2.7: Schematic of Angular Potentiometer Operation [11]

central coil which induces voltages in each of the secondary coils. These induced voltages are linearly related

to the position of the magnetic core which is attached to the member whose displacement is of interest [28].

As the magnetic core moves, the changing voltages in the secondary coils are converted to the amount of

displacement [11]. LVDT sensors are generally more expensive than other displacement options, but they

provide significant advantages. This type of sensor is capable of measuring dynamic displacements, can

measure less than a nanometer of displacement, are suitable for use in very low temperatures with minimal

loss in accuracy, and they also are virtually immune to wear damage because the moving magnetic core

makes no physical contact with the electrical coils [11, 29, 28]. LVDT displacement sensors can be costly,

but are applicable in many situations where linear potentiometers cannot be utilized.

Figure 2.8: Linear Variable Differential Transformer Internal Composition Diagram [11]

21
2.4.4 Tiltmeters and Inclinometers

As mentioned previously, another metric that may be of interest in structural health monitoring is the

slope or tilt of a particular location. Tiltmeters and inclinometers are the sensors utilized to measure this

parameter. Three different options are available for tiltmeters each with their own benefits. These will be

compared in the following paragraphs.

Vibrating Wire Tiltmeters

Vibrating wire tiltmeters are composed of a pendulous mass which is supported by a vibrating wire

strain gage. As the sensor tilts, the gravitational force from the mass upon the strain gage will change

due to this rotation. The frequency of the wire within the strain gage will shift because of this, and the

relationship between this frequency and the angle of rotation can be determined. These sensors are designed

for permanent measurements and are not applicable for dynamic slope changes, such as those due to traffic

loads or an earthquake [11].

Electrolytic Tiltmeters

Electrolytic tiltmeters utilize a high-precision electrolytic tilt transducer and the primary sensing element.

These devices are excited by an AC signal through the tilt transducer, and as the transducer tilts, internal

electrodes are covered or uncovered by a conductive fluid within the device. As these electrodes are exposed

or removed from the conductive fluid, the electrical resistance to the excitation signal is altered which can be

converted to a change in angle through a scaling factor. The primary benefit to this sensor type is that the

angular movement of the sensor and thus the member it is measuring, is compared to the gravity vector and

therefore no external datum is necessary for operation. These sensors again are best for long-term angular

changes rather than rapidly changing tilts [11].

Inertial Based Inclinometers

Inertial based inclinometers are the main option for measuring dynamic tilt changes. The technology in

these devices is similar to that of a servo force balance accelerometer. These inclinometers also contain a

pendulous mass, and the motion of this mass is detected by an optical position sensor. The reading from

this position sensor is compared to the original state and a balancing current is produced and applied to

a torque motor to return the mass to the original state. This basic operation is portrayed in a diagram in

Figure 2.9. The balancing current required to keep the mass in the initial position is proportional to the tilt

22
experienced by the sensor. These inclinometers are very accurate with the ability to record changes in slope

as minute as 0.1 seconds of a degree and are applicable where high levels of shock and vibration occur [11].

Figure 2.9: Diagram of the Operation of an Inertial Based Inclinometer [30]

2.4.5 Acoustic Emission Sensors

The most common type of acoustic emission sensor uses piezoelectric materials similar to the operation

of a piezoelectric accelerometer. This type of sensor has been proven to be more durable and sensitive than

other techniques based upon capacitance or laser-optics [31]. The piezoelectric material within the sensor

experiences the force from the stress waves propagating from a crack and outputs a proportional electrical

signal. A basic diagram of an acoustic emission sensor is shown in Figure 2.10. In order for these sensors

to be applicable, they are typically used within a network of multiple sensors so that triangulation can be

performed to determine the location of a stress wave source [11]. The best advantages of acoustic emission

sensors is the ability to detect damage as it occurs and the fact that the damage does not need to occur at

the site of the sensor for it to be detected. Steel is the best material for the use of acoustic emission sensors

because the stress waves do not attenuate as drastically and can be detected at farther distances [10].

23
 Figure 2.10: Diagram of an Acoustic Emission Sensor [31]

2.4.6 Fiber Optic Sensors

Fiber optic sensors are a relatively new technology that have a lot of promising applications. The basic

operation of these sensors involves sending light beams down a fiber optic cable at regular intervals and

measuring the changes that occur as this light is reflected [12]. Four different principles are commonly used

to interpret these changes in light signals: interferometry, polarization, spectroscopy, and light intensity [32].

Interferometry based fiber optic sensors monitor changes in the shape of light waves to measure environmental

changes. When the principle of polarization is employed, the changes in optical polarization of a light signal

are the measure of interest [33]. One of the most common types of fiber optic sensor types is a fiber Bragg

grating (FBG) sensor which is based on the principle of spectroscopy. Spectrally based sensors like FBG

sensors monitor the change in wavelength of reflected light sources. Sensors using the principle of light

intensity measure how light intensity changes over the length of a fiber [10].

There are a large number of advantages to the use of fiber optic sensors which has fueled the amount

of research behind these devices. The most significant benefit to the use of fiber optics is the ability to

measure multiple parameters with only one fiber because of the variety of principles that can be employed

[11, 10, 12]. Another notable advantage over other types of sensors is the capability to measure changes

along the entire length of a fiber rather than just at one particular location. Fiber optic sensors are also

immune to electromagnetic interference which is a common problem with multiple other sensor options.

24
Fiber optic technology can be easily incorporated into other devices such as accelerometers and electrical

transducers and have been used to measure a variety of metrics including strain, displacement, pressure,

slope, acceleration, corrosion, loading, and cracking in concrete [32, 34]. Fiber optic strain gages that have

been developed have shown the ability to detect strains with several orders of magnitude improvement over

common electrical resistance strain gages and are immune to any transverse effects which other strain gages

are often affected by [11, 12]. Two models of fiber optic strain gages are common: fiber Bragg gratings (FBG)

sensors and extrinsic Fabry-Perot interferometric (EFPI) sensors. FBG sensors detect how the wavelength

of reflected light changes due to applied loads. These sensors can be very small because the light changes

are measured over a pattern written on the optical fiber which can be as short as needed. Fiber Bragg

based strain gages are less expensive that EFPI sensors, but are not as accurate. In EFPI gages, the sensing

element is not the optical fiber itself; instead it consists of two mirrors placed that the tips of optical fibers

and the wavelength of reflected light between them is a function of the distance. EFPI strain gages are the

most sensitive option, but are also the most expensive [11]. The leading drawback to the use of fiber optic

sensors comes from this higher cost which has been a major obstacle in their process of becoming commonly

utilized [11]. Further research and development are constantly being conducted in the field of fiber optic

sensors for structural health monitoring to discover new applications and increase the affordability so that

their distinct advantages may be more available in the future.

2.4.7 Microelectromechanical Systems (MEMS) Sensors

Sensors utilizing microelectromechanical systems (MEMS) are another recent technology development in

the field of structural health monitoring. One of the significant advantages of MEMS sensors is the ability to

both sense and actuate. This means that within the same device data can be collected and partially processed

before being transmitted. This is usually done by including an on-board microprocessor within the sensor

system which can be used to convert the signal to digital, perform basic calculations, or provide interfacing

functions which can greatly reduce the amount of data processing required [12]. MEMS sensors also have

the benefit of miniature size, so they are applicable in situations where typical sensors are too large. The

common manufacturing process for MEMS devices also presents the possibility of large scale productions

with relatively low cost [35]. One application found in research is a MEMS-based transducer to be used

for acoustic emission detection. This type of acoustic emission sensor could include multiple transducers,

each with a narrow and highly sensitive resonant frequency, which together still cover the frequency range of

interest. This could help the process of differentiating environmental noise from actual stress waves caused

25
by acoustic emission events and greatly improve the accuracy overall [35]. MEMS-based strain sensors have

also been developed using piezoresistive principles which could improve resolution and sensitivity as well as

consume less energy than common strain gages [36].

MEMS sensors also have a great deal of potential for use outside of structural health monitoring. One

of the promising applications is the use of MEMS sensors in microbiology. MEMS sensors have been shown

to be ideal for measurement of force and displacement on the single-cell level with the ability to study a

large population of cells at one time [37]. This ability has been utilized in tests to recognize and remove

cervical cancer cells from a mixed cell population. The hope is for these sensors to be a diagnostic tool

because they can process a large population rapidly with low cost [38]. A potential new MEMS-based sensor

with possible applications in both structural health monitoring and microbiology is being developed and

preliminary calculations and investigations for this device are discussed in Chapter 4.

2.5 Summarizing Remarks

The field of structural health monitoring has grown significantly over the last few decades. The aging

infrastructure in the United States and around the world as well as rapidly improving technologies have

helped this growth. Multiple methods have been investigated and tested to assess the overall state of a

structure through global health monitoring as well as concepts to determine if issues are present on a more

local level. Numerous varieties of sensors and devices have also been developed to accurately gain the

data required for global and local assessment practices, with some very successful and applicable options

available. Even with all of this progress, there still exists considerable issues and complications with many

of the methods and technologies that have been developed. Although an increasing number of bridges and

other structures are being studied through structural health monitoring concepts, a lot more research and

development need to be performed before measurement and investigation through these methods can be

feasible for all structures. The goal to continually improve the accuracy, applicability, and cost effectiveness

of structural health monitoring is the impetus behind the development of the MEMS device investigated in

Chapter 4 and many similar current projects and ideas for the future. Through all of this continued research,

the achievement of monitoring all bridges and structures may someday be met so that tragedies such as the

I-35W bridge collapse would never again occur.

26
Chapter 3

Theory of Beam Vibration

3.1 Introduction

The vibrating beam is a well studied mechanics problem with a multitude of practical applications in

engineering. Problems involving a vibrating beam are typically approached using one of the common beam

theories, which were all introduced by 1921 [39] and have been thoroughly studied. The three most typically

used beam theories are the Euler-Bernoulli, Rayleigh, and Timoshenko theories. While for many engineering

applications, the use of one of these theories provides reasonable results, all three of them still rely upon a

few basic assumptions. These assumptions take fully three-dimensional elasticity theory and simplify it to an

approximate one-dimensional theory. A main component that is missing from the collection of research on the

vibrating beam is an investigation into when these assumptions begin to fall apart. In order to confidently

utilize these beam theories their limitations should be well understood. In this chapter a fourth method

to analyze vibrational beam mechanics is introduced that better approximates a fully three-dimensional

elasticity theory. This approach will be employed to study the accuracy of the common beam theories and

the point at which they fail to be acceptable analysis tools. When studying the effectiveness of common

beam theories, three different boundary condition cases will be evaluated: fixed-free, fixed-fixed, and simply

supported. The following sections will discuss the concepts applied in each theory and the development of

equations of motion and boundary conditions for each. Frequency equations for each of the proposed end

condition cases will also be introduced and explained.

27
3.2 Basic Beam Theories in Mechanics

3.2.1 Euler-Bernoulli Beam Theory

The Euler-Bernoulli beam theory, also known as the classical beam theory, Euler beam theory, Bernoulli

beam theory, or Bernoulli-Euler beam theory, is the most commonly used theory. The early development

of this theory can be traced back to the early 18th century. Two different Bernoulli’s contributed to this

model, but it began with Jacob Bernoulli (1654-1705). Jacob was the first to use infinitesimal calculus

to study the deflection curve of an elastic bar. He developed an equation that described a proportional

relationship between the bending moment at a certain point in a beam and the curvature at that location

[40]. Although the constants he found in this equation were later found to be incorrect, the basic principle was

true and was later used by Euler. Leonard Euler (1707-1783) made significant advancements to mathematics

and mechanics of materials, and he happened to be the pupil of another Bernoulli, Jacob’s nephew, Daniel.

Daniel Bernoulli (1700-1782) contributed to this theory in two ways. He was the first to derive the differential

equation that governs the lateral vibrations of a prismatic bar and also made a crucial suggestion that Euler

use variational calculus in deriving equations for the deflection curves of elastic beams. Euler took this

suggestion, and also adopted Jacob Bernoulli’s principle, when he expanded the study of beam mechanics.

Euler showed how differential equations of motion could be derived and how the motion of a particle could

be found by integrating these equations. He used these principles to establish formulae for the frequencies

and mode shapes of beams with varying end conditions [40]. The final equations and principles described in

Euler’s work contrive the final form of the Euler-Bernoulli beam model.

This beam model is the most commonly used theory because it is very simple and provides acceptable

results for many engineering problems. This theory includes both the strain energy from bending and the

kinetic energy from transverse displacement, but does not consider the effects of rotary inertia or shear

displacement. Due to ignoring these two contributions, this model tends to overestimate the natural fre-

quencies of a vibrating beam, especially for higher modes of vibration [39]. The Euler-Bernoulli model also

lacks accuracy as the beam becomes less slender. When analyzing low modes of vibration for slender beams,

the contributions from rotary inertia and the effect of shear are negligible, but this does not describe many

engineering problems. Therefore, more advanced beam models have been developed to improve the results

for less slender beams and higher modes of vibration.

28
3.2.2 Rayleigh Beam Theory

It was almost a century after the work of Euler before a new contribution was made to beam mechanics

theories. This improvement came from Lord Rayleigh (1842-1919) in his treatise “The Theory of Sound” in

1877. In this work, Rayleigh proposed a method of finding frequencies of vibration directly from an energy

view by assuming a suitable form for the type of motion, rather than solving differential equations. He then

used this approximate method to study transverse vibrations of beams. His most important contribution

from this study was a correction that allowed for rotary inertia which had been ignored before [40]. When

the equations of motion that include this effect of the rotation of the cross section are employed, it is called

the Rayleigh beam theory. The addition of this factor provides a slight improvement over the Euler-Bernoulli

model by slightly reducing the overestimation of the natural frequencies. The Rayleigh model still does not

incorporate the contribution of shear displacement, and thus does not entirely fix the high frequency results

[39].

3.2.3 Timoshenko Beam Theory

In 1921 Stephen Timoshenko (1878-1972) proposed a beam theory that presents significant advancements

over the Rayleigh beam theory. The Timoshenko beam theory, sometimes known as the first order shear

deformation theory, expands upon the Euler-Bernoulli theory by incorporating the effects of both rotary

inertia and shear deformation. In his calculations, Timoshenko found that the change due to the addition of

shear was four times greater than the change from the inclusion of rotary inertia [41]. Because of this, the

Timoshenko beam theory is more popular than the Rayleigh model. This model results in far more accurate

frequency values for non-slender beams and high frequency modes than either of the other two options. In

these two situations, the effect of shear forces on beam deflections is significant and cannot be neglected [39].

Therefore, the use of the Timoshenko model has significant applications beyond the potential use of either

the Euler-Bernoulli or Rayleigh beam theories. An essential component in the Timoshenko model is the

shape factor. This parameter is utilized to account for the variation of shear stress across the cross section.

The shape factor is a function of the shape of the cross-section, Poisson’s ratio, and the frequency of vibration

[39]. Usually, the dependence on the frequency is ignored, leading to less accurate results. In general, the

Timoshenko model provides a substantial improvement over both the Euler-Bernoulli and Rayleigh beam

theories and is still widely used.

29
3.2.4 Elasticity Based Beam Theory

In order to study the accuracy of typical beam theories a fully three-dimensional elasticity approach is

proposed. A more advanced theory is considered because all three of the discussed theories still rely on a

number of basic assumptions [39]. These include:

• The axial dimension is significantly larger than the other two.

• The Poisson effect is neglected.

• Planes perpendicular to the neutral axis remain perpendicular after deformation.

• The material is isotropic.

• The material is linear elastic.

• The cross-section is symmetric.

• The angle of rotation is small enough that the small angle approximation can be used.

The proposed elasticity theory will remove a few of the assumptions listed above. The problem of the

transversely vibrating beam typically cannot be solved exactly, so solutions using Ritz-based approximations

are utilized in this advanced method. This allows for the inclusion of the Poisson effect and the ability to

calculate accurate frequencies for non-slender beams and even anisotropic materials. The details of this fully

three-dimensional elasticity method are described in the following section.

3.3 Governing Beam Theory Equations

Based on the assumptions made and the different mechanical effects that are included, each beam model

has unique governing equations associated with it. These equations represent the principles included in each

model and the motion of the beam under loading or vibrations. A few different categories of equations are

presented and explained for each model in this section.

A key tool employed in the analysis of beam response is the assumption of a displacement field based

on certain kinematic restraints imposed by the model. The general purpose of restraining the motion of a

particle through a displacement field representation is to minimize the number of dimensions in a problem.

By limiting the dimensions considered in analysis, the computation can be significantly simplified. As the

theories progress from the simplest representation, Euler-Bernoulli Theory, to the most complex, Elasticity

30
Theory, the constraints on the motion are reduced leading to more complex displacement fields. The assumed

displacement field for each model discussed in this chapter will be introduced in the following paragraphs.

Another common method to characterize the response of beam uses the governing differential equation

or equations of motion for a particular model. This is generally developed using Hamilton’s principle [39].

This principle is based upon the Lagrangian function, sometimes called L, which is given as

L ≡ U − K + A. (3.1)

In this function, U represents the total strain energy of the beam, K is the total kinetic energy of the beam,

and A represents the potential energy of the loading [42]. For the study discussed in this paper, only free

vibration is considered and the loading term A drops out. The equation for Hamilton’s principle can be

expressed as
Z t1
δ Ldt = 0. (3.2)
t0

This principle states that:

The time integral of the Lagrangian function over a time interval t0 to t1 is an extremum for

the “actual” motion with respect to all admissible virtual displacements which vanish, first, at

instants of time t0 and t1 at all points of the body, and, second over Su , where the displacements

are prescribed, throughout the entire time interval [42].

The Lagrangian function can be found for each of the beam theories discussed based upon the assumed

displacement field and the different displacement effects which are included. By employing this principle,

the governing differential equation of motion can be found for each theory. Along with the equation of

motion, the boundary conditions that must be satisfied are also developed using this principle. Using the

developed equations of motion, formulas for the spatial solution to each problem have been found. The

differential equations of motion, boundary conditions, and spatial solutions that have been developed for

the three typical beam theories will be discussed in this section. For the full elasticity theory, a generalized

eigenvalue problem is developed rather than a differential equation to represent the motion of a beam, which

will also be presented in this section.

The same variables will be used for the discussion of all four models. In accordance with common practice

for beam theory discussion, dimensionless variables will be employed for the geometric and vibrational

parameters. The notation used follows

31
 • x1 = x = the dimension parallel to the width of the beam

• x2 = y = the vertical or transverse dimension

• x3 = z = the axial dimension

• u1 = u = the unknown displacement in the x-direction

• u2 = v = the unknown vertical or transverse displacement

• u3 = w = the unknown axial displacement

• L∗ = beam length

• A∗ = cross section area

• I ∗ = moment of inertia

• ρ∗ = material density

• E = Youngs Modulus

• G = Shear Modulus

• ωi∗ = angular frequency

• L = L∗ /L∗ = dimensionless beam length

• A = A∗ /L∗2 = dimensionless area

• I = I ∗ /L∗4 = dimensionless moment of inertia

• ρ = ρ∗ L∗6 ω1∗2 /(EI ∗ ) = dimensionless density

• ωi = ωi∗ /ω1∗ = dimensionless angular frequency

3.3.1 Euler Bernoulli Model

Displacement Field

The Euler-Bernoulli beam theory contains only one unknown variable, the transverse displacement of

the beam centroid. This displacement is only a function of the axial dimension, which is the reason this

32
is considered a one-dimensional theory. Since the Poisson effect is neglected in this model, the assumed

displacement field is of the form

u(x, y, z) = 0

v(x, y, z) = v(z)

dv
w(x, y, z) = −y .
dz

Equation of Motion and Boundary Conditions

The Lagrangian function determined from the kinetic and potential energy terms included in the Euler-

Bernoulli model is given by [39]

" 2 2 #
1
∂ 2 v(z, t)
Z  
1 ∂v(z, t)
L= ρA − dz. (3.3)
2 0 ∂t ∂z 2

When this function is substituted into Hamilton’s principle the equation of motion and boundary conditions

are determined. This governing differential equation of motion is given by [39]

∂ 2 v(z, t) ∂ 4 v(z, t)
ρA + = f (z, t). (3.4)
∂t2 ∂z 4

As mentioned before, only free vibration is considered for the vibrational study in this paper, so the forcing

function, f (z, t), is set to zero. The boundary conditions which need to be satisfied for this equation to be

applicable are [39]

1 1
∂2v ∂3v
 
∂v
δ =0 δv = 0. (3.5)
∂z 2 ∂z 0 ∂z 3 0

Each of the terms in these boundary conditions have a physical meaning that they represent. The expression

v represents the transverse displacement, and thus the first derivative ∂v/∂z represents the slope. The second

derivative of the displacement ∂ 2 v/∂z 2 represents the moment, and finally the third derivative ∂ 3 v/∂z 3 is

the shear force. The δ operator signifies the variation of the term it is operating on. Therefore, if δv = 0

it means that the displacement is constant, or known. It does not necessarily mean that the value of the

displacement is zero. Similarly, the expression δ(∂v/∂z) = 0 indicates that the slope is known [39].

There are four different possible support conditions that can be considered for a beam. These are: a free

end where the displacement and slope are unknown, a fixed end where the moment and shear are unknown,

33
a hinged end where the shear and slope are unknown, and a rolling end where the displacement and moment

are unknown. Each of these four possibilities have their own respective boundary condition combinations

which must be satisfied. These are listed below.

∂2v ∂3v
• for a free end: ∂z 2 = 0 and ∂z 3 =0

∂v
• for a fixed end: v = 0 and ∂z =0

∂2v
• for a hinged end: v = 0 and ∂z 2 =0

∂v ∂3v
• for a rolling end: ∂z = 0 and ∂z 3 =0

These known boundary conditions will be utilized to find mode shape and frequency equations for the varying

beam cases that will be studied.

Spatial Solution

To simplify the analysis of different beam conditions, the principle of separation of variables has been

used to isolate an equation for transverse displacement that is dependent on the axial direction only, and

not time [39]. This new equation is referred to as the spatial solution and it is of the form

V (z) = C1 sin az + C2 cos az + C3 sinh az + C4 cosh az. (3.6)

In this equation, the parameter a is the dimensionless wave number. This value represents 1/2π times the

number of wave cycles in a beam length [39]. For the Euler-Bernoulli model this number is given by

a = ρAω 2 . (3.7)

Since the spatial solution equation listed above is not dependent upon time, the boundary condition equations

need to be altered before their substitution into the spatial solution. The terms in the spatial boundary

conditions become ordinary derivative, rather than partial and the variable v(z, t) is replaced with V (z).

These updated boundary conditions can be substituted into the spatial solution, resulting in a system of

equations. This system of equations can be solved to find the unknown C factors and equations for the

vibrational frequency of different beam cases. These will be presented in a later section.

34
3.3.2 Rayleigh Beam Theory

Displacement Field

The assumed displacement field for the Rayleigh beam model is identical to the Euler-Bernoulli model.

The inclusion of rotational inertia effects alters the kinetic energy of the beam, but does not affect the

assumed displacement relationships. The model still only has the one unknown variable of the transverse

displacement and ignores the Poisson effect. Therefore the displacement field is still of the form

u(x, y, z) = 0

v(x, y, z) = v(z)

dv
w(x, y, z) = −y .
dz

Equations of Motion and Boundary Conditions

The Lagrangian function for the Rayleigh beam theory includes the term for the kinetic energy due to

the rotation of the cross section [39]. With this additional term the Lagrangian becomes
" 2 2 2 #
1
∂ 2 v(z, t) ∂ 2 v(z, t)
Z   
1 ∂v(z, t)
L= ρA + ρI − dz. (3.8)
2 0 ∂t ∂t∂z ∂z 2

The addition of this kinetic energy term from the rotation alters the governing differential equation of motion

found using Hamilton’s principle for the Rayleigh beam model. With this factor included the form the of

equation becomes [39]

∂ 2 v(z, t) ∂ 4 v(z, t) ∂ 4 v(z, t)
ρA 2
+ 4
− ρI = f (z, t). (3.9)
∂t ∂z ∂z 2 ∂t2

Similar to the Euler-Bernoulli model, the function f (z, t) is set to zero to represent free vibration. The

boundary conditions that need to be satisfied for this model are [39]

1 1
∂2v ∂3v ∂3v
   
∂v
δ =0 − ρI δv = 0. (3.10)
∂z 2 ∂z 0 ∂z 3 ∂z∂t2 0

The boundary condition terms that represent the displacement, slope, and moment are the same as in the


Euler-Bernoulli model. The term that represents the shear force has changed to become ∂ 3 v/∂z 3 − ρI∂ 3 v/∂z∂t2 .

Since this boundary term now includes a time derivative, it will have to be redefined to be applicable with

35
the spatial solution. This is discussed in the following paragraph. Again, there are four possible support

conditions, each with unique boundary condition combinations. These are listed below.
 
∂2v ∂3v ∂ v3
• for a free end: ∂z 2 = 0 and ∂z 3 − ρI ∂z∂t 2 =0

∂v
• for a fixed end: v = 0 and ∂z =0

∂2v
• for a hinged end: v = 0 and ∂z 2 =0
 
∂v ∂3v ∂ v3
• for a rolling end: ∂z = 0 and ∂z 3 − ρI ∂z∂t 2 =0

Spatial Solution

By separating the time and spatial variables in the governing equation of motion, the spatial solution for

the Rayleigh beam theory was found to be [39]

V (z) = C1 sin az + C2 cos az + C3 sinh bz + C4 cosh bz. (3.11)

In this model there are two dimensionless wave numbers, which are defined by the following expressions [39].

v s
u 2
u ρIω 2 ρIω 2
a= + + ρAω 2 (3.12)
t
2 2
v s
u 2
u ρIω 2 ρIω 2
b= − + + ρAω 2 (3.13)
t
2 2

The boundary conditions listed in the previous section need to updated to be compatible with the spatial

solution. For the displacement, slope, and moment conditions the partial derivatives become ordinary

derivatives and the v(z, t) variable is changed to V (z) similar to in the Euler-Bernoulli model. The shear

term in this case cannot be updated as simply as before though, since it includes a spatial derivative. In this

case, the expression for shear in the boundary terms becomes [39]

d3 V dV
+ ρIω 2 .
dz 3 dz

These spatial boundary conditions can be applied to the spatial solution to determine the unknown C

constants and frequency equations for the Rayleigh model. These solutions will be discussed in a following

section.

36
3.3.3 Timoshenko Beam Theory

Displacement Field

Since the Timoshenko beam theory includes deformation due to shear, the total angle of rotation for a

cross section is now the sum of both the bending and shear deflection effects. Thus, the Timoshenko model

increases the number of unknown variables to two. The total cross sectional rotation due to shear is now

unknown as well as the transverse displacement. In the spatial description the variable for total rotation

is Ψ(z). This variable is only dependent upon the axial dimension, similar to the transverse displacement.

The Timoshenko beam theory is therefore still only a one dimensional theory. Since it is one-dimensional,

the Poisson effect is again neglected and the assumed displacement field becomes

u(x, y, z) = 0

v(x, y, z) = v(z)

w(x, y, z) = −yΨ(z).

Equations of Motion and Boundary Conditions

If strain-displacement and constitutive laws are used with the above displacement field to determine

the values for stress, a significant problem is encountered. Finding stress through this method results in a

constant shear stress throughout the cross section. This is physically impossible since the outside surfaces

of the beam are traction free. Timoshenko realized this issue and introduced the shear factor which was

discussed before. This factor partially corrects this issue, but is still somewhat approximate. For the purposes

of the vibrational study in this paper, a shear factor that is dependent on the cross section shape and Poisson

ratio will be used for calculations involving an isotropic material. For a rectangular cross-section the shape

factor is given as [39]

10(1 + ν)
k0 = . (3.14)
12 + 11ν

For an orthotropic material, determining an adequate shear factor is not so simple. Instead of the equation

listed above, a new shear correction factor was determined using Hutchinson’s beam theory by Puchegger

and co-workers [43]. The shear factor found through this method is related to the cross section shape, two

Poisson ratios, two shear moduli, and Young’s modulus in the axial direction.

37
 The derived equation is given by

Ez
k0 = − . (3.15)
Gzy [(A/Ix2 )C4 + νzy − (Iy /Ix )νzx ]

In this equation C4 is a constant that can be found using the following formula.

f1 (x, y)2 f2 (x, y)2

Z  
C4 = [−y 2 νzy f1 (x, y) + x2 νxz f1 (x, y) − 2yxνxz f2 (x, y) − Ez + ]dxdy (3.16)
A Gzy Gxz

The functions f1 and f2 can be found in the paper by Puchegger and others [43].

By using these shear factors, the potential energy from shear can be included in the Timoshenko beam

model. The Lagrangian function is therefore given by [39]

" 2 2 2 2 #
1
∂ 2 v(z, t)
Z    
1 ∂v(z, t) ∂α(z, t) ∂v(z, t)
L= ρA + ρI − − k 0 GA − α(z, t) dz. (3.17)
2 0 ∂t ∂t∂z ∂z ∂z

Since there are two unknown variables for this theory, using Hamilton’s principle produces two differential

equations of motion which are [39]

∂ 2 v(z, t)
 2 
0 ∂ v(z, t) ∂α(z, t)
ρA − k GA − = f (z, t)
∂t2 ∂z 2 ∂z
∂ 2 α(z, t) ∂ 2 α(z, t)
 
0 ∂v(z, t)
ρI − − k GA − α(z, t) = 0. (3.18)
∂t2 ∂z 2 ∂z

The function f (z, t) is set to zero again because only free vibration is considered for the analysis. The

boundary conditions which need to be satisfied for these equations are given by

1   1
∂α ∂v
δα =0 k 0 GA − α δv = 0, (3.19)
∂z 0 ∂z 0

where α(z, t) represents the total rotation of the cross section [39].

The boundary condition terms have a different appearance in the Timoshenko model because of the

additional variable of section rotation, but the physical meaning of the terms is very similar to that of the

Euler-Bernoulli and Rayleigh beam theories. The expression for the transverse displacement is still v, but the

term for the slope is now simply represented by α rather than the derivative of the displacement. Similarly,

the moment term is now ∂α/∂z rather than the second derivative of transverse displacement. The shear

term in this model is quite different from the previous two theories since the shear factor and shear modulus

38
are now included. The expression for shear is now k 0 GA(∂v/∂z − α). With these new boundary condition

expressions, the combinations for each of the four possible support conditions have changed and are listed

below.

∂α
= 0 and k 0 GA ∂v


• for a free end: ∂z ∂z −α =0

• for a fixed end: v = 0 and α = 0

∂α
• for a hinged end: v = 0 and ∂z =0

• for a rolling end: α = 0 and k 0 GA ∂v

∂z −α =0

Spatial Solution

In order to determine spatial solutions for both the transverse displacement and the total rotation the

governing equations provided in the previous section must first be decoupled so each equation only contains

one of the unknown variables [39]. The decoupled form of these equations is given below.

∂4v  ρ  ∂4v ∂2v ρ2 I ∂ 4 v

4
− ρI + 0 2 2
+ ρA 2 + 0 =0
∂z k G ∂z ∂t ∂t k G ∂t4
4 4 2 2 4
∂ α  ρ  ∂ α ∂ α ρ I∂ α
4
− ρI + 0 2 2
+ ρA 2 + 0 =0 (3.20)
∂z k G ∂z ∂t ∂t k G ∂t4

For these equations, when the method of separation of variables is applied it is found that two different

cases need to be considered for the spatial solutions [39]. These two cases are when the frequency is less

than a critical value, or if it is above this value. In Han and co-workers work, it was shown that this critical
p
frequency value is equal to k 0 GA/ρI and is referred to as ωc .

For the case when ω < ωc the form of the spatial solutions for the transverse displacement and total

section rotation are given by

         
 V (z)   C1   C2   C3   C4 
 =  sin az +   cos az +   sinh bz +   cosh bz. (3.21)
Ψ(z) D1 D2 D3 D4

The expressions for the two dimensionless wave numbers for this case are given by the following expressions

[39]. v
u s 2
ρω 2 ρ2 ω 4

u 1 1
a= I+ 0 + I− + ρAω 2 (3.22)
t
kG 2 k0 G 4

39
 v
u   2 s 2 2 4
u 1 ρω 1 ρ ω
b= − I+ 0 + I− 0 + ρAω 2 (3.23)
t
kG 2 kG 4

Although it seems that there are now eight unknown constants to solve for the Timoshenko beam model,

the Ci and Di values are related to each other and thus only four unknowns need to be determined for the

full solution [39]. The relationships between these factors are given below.

k 0 GAa2 + ρAω 2 k 0 GAa2 − ρAω 2

D1 = − C2 D2 = C1
k 0 GAa k 0 GAa
k 0 GAb2 + ρAω 2 k 0 GAb2 + ρAω 2
D3 = C4 D4 = C3 (3.24)
k 0 GAb k 0 GAb

The equations are slightly different for the case when ω > ωc . The spatial solutions become [39]

         
V (z) C¯1 C¯2 C¯3 C¯4
=  sin az +   cos az +   sinh b̄z +   cosh b̄z. (3.25)
         

Ψ(z) D̄1 D̄2 D̄3 D̄4

The expression for the dimensionless wave number, a, is the same as for the previous case. The equation for

b̄ is slightly different, though, and is given by

v s
u 2
ρω 2 ρ2 ω 4

u 1 1
b̄ = I+ 0 − I− + ρAω 2 . (3.26)
t
kG 2 0
kG 4

The constants C̄i and D̄i are also interrelated for the case of ω > ωc , so there are still only four unknowns

for this case. These relationships are expressed as [39]

k 0 GAa2 + ρAω 2 ¯ k 0 GAa2 − ρAω 2 ¯

D̄1 = − C2 D̄2 = C1
k 0 GAa k 0 GAa
k 0 GAb̄2 + ρAω 2 ¯ k 0 GAb̄2 − ρAω 2 ¯
D̄3 = − C4 D̄4 = − C3 . (3.27)
k 0 GAb̄ k 0 GAb̄

The dimensional wave numbers b and b̄ are also related through the following equation.

b = ib̄ (3.28)

All of these relationships, along with the spatial solutions can be combined with known boundary con-

ditions for varying support conditions to find the mode shapes and frequency equations for different beam

cases. The boundary conditions provided in the previous section need to be updated to be compatible with

40
the spatial solution, similar to the Euler-Bernoulli and Rayleigh models. This is simply done for the Timo-

shenko model by changing partial derivatives to ordinary, replacing α(z, t) with Ψ(z), and replacing v(z, t)

with V (z).

3.3.4 Linear Elasticity Theory

The formulation for the mode shape and frequency solutions through a three-dimensional linear elasticity

theory differs significantly from the three one-dimensional theories previously discussed. Rather than using an

assumed displacement field, governing differential equations and spatial solutions, Ritz-based approximations

are used to solve Hamilton’s principle. The details of this are presented in the following paragraphs.

Constitutive Relationship

In order to consider all three dimensions including the Poisson effect, the development of this theory

begins with the generalized Hooke’s Law. This constitutive relationship in indicial form is given as

σij = Cijkl kl (3.29)

where σij represents the stress tensor, Cijkl is the elastic stiffness tensor, and kl is the strain tensor. In this

full form there are 9 stress and strain components and 81 stiffness terms. This would pose a very complicated

problem, but fortunately the problem can be significantly simplified. Using the principle of symmetry it can

be shown that σij = σji and kl = lk . This reduces the problem to 6 stress and strain components and

only 36 stiffness terms. It has also been proven that the stiffness matrix is symmetric which reduces this

number even further to 21 constants. Lastly, if the considered material is orthotropic and has three planes

of symmetry the stiffness matrix only contains nine independent constants. This reduced constitutive law

that will be considered in analysis can be expressed as follows.

    



 σ11 


  C11 C12 C13 0 0 0 


 11 




 
   

   


 σ22 


 C
 12 C22 C23 0 0 0 
 22 



 
   

   
 σ33
 
  C13

C23 C33 0 0 0 

33 

=


 (3.30)
σ23  0 0 0 C44 0 0 γ23 

 
  

 
   


 
   

   



 σ13 



 0
 0 0 0 C55 0 

 γ13 




 
   

 σ12
 
 0 0 0 0 0 C66

 γ12 


41
The variable γ in this equation represents physical shear stress which is equivalent to 2. The elastic stiffness

constants in this matrix equation are related to the basic material properties E, G, and ν through the

following equations.

E(1 − ν)
C11 = C22 = C33 = (3.31)
(1 + ν)(1 − 2ν)
Eν
C12 = C13 = C23 = (3.32)
(1 + ν)(1 − 2ν)
C44 = C55 = C66 = G (3.33)

This constitutive law will be utilized in the solution for the elasticity-based approximation.

Hamilton’s Principle

For this full elasticity theory Hamilton’s principle can be written as

Z tZ
0=− {σ1 δ1 + σ2 δ2 + σ3 δ3 + σ4 δ4 + σ5 δ5 + σ6 δ6 }dV dt
0 V
Z tZ
1
+ δ ρ(u̇2 + v̇ 2 + ẇ2 )dV dt (3.34)
2 0 V

where u̇ = ∂u/∂t and the conventional notation for stress (σ11 = σ1 , σ23 = σ4 , etc.) has been used [44].

For this advanced beam model, selected displacement approximation functions will be used to find a

solution to the above equation, rather than using potential and kinetic energy equations based upon an

assumed one-dimensional displacement field and a few included displacement effects. Approximations are

used because of the difficulty in finding an exact solution to the free vibration of a three-dimensional solid.

The above equation only contains displacement variables in the second integral, and only stress and strain

terms are found in the first integral. To determine expressions for these stress and strain terms, strain-

displacement relationships are utilized as well as the constitutive law discussed previously. The basic strain-

displacement equation that is utilized for this analysis for small displacements is given as

 
1 ∂ui ∂uj
ij = + . (3.35)
2 ∂xj ∂xi

The strains found from this relation are used as the δ terms in Hamilton’s equation. These strain expressions

can then be substituted into the constitutive law to determine the stress expressions. For this study, Ritz-

based approximations will be utilized for the displacement functions and are explained in the next section.

42
Ritz Approximations

The namesake for the Ritz method comes from Walther Ritz who was a Swiss theoretical physicist.

When Ritz was working with the bending of rectangular plates, he proposed approximation functions in the

form of a finite series [40]. When used for displacement the general form for the three direction component

approximations can be written as

l
X
u(x, y, z, t) = φuo (x, y, z) + cj (t)φuj (x, y, z)
j=1
m
X
v(x, y, z, t) = φvo (x, y, z) + dj (t)φvj (x, y, z)
j=1
n
X
w(x, y, z, t) = φvo (x, y, z) + ej (t)φw
j (x, y, z). (3.36)
j=1

In these equations c, d, and e are unknown constants that depend on time. The φo terms represent the

simplest functions that satisfy the essential boundary conditions for that displacement direction. In the study

discussed in this paper, the initial boundary conditions are assumed to be zero, with no initial displacement

or velocity. The φo terms are therefore all equal to zero for the considered case. In these approximation

equations, φj represents a selected approximate function for each respective direction. These functions must

fulfill the three following requirements.

1. They satisfy the homogeneous form of the essential boundary conditions.

2. They are linearly independent of each other.

3. They must become increasingly more complex from the first term to the last used.

These Ritz-based approximations will be substituted into the equation for Hamilton’s principle presented

before to calculate a spatial solution. By using a large number of approximation terms for each displacement

component, very accurate solutions can be determined for the mode shapes and frequencies of vibration for

a beam. When the substitution into Hamilton’s principle equation is done and harmonic motion is assumed,

the problem is reduced to a generalized eigenvalue problem expressed as follows.

     
11
 [K ] [K 12 ] [K 13 ]  
 {c} 





11
 [M ] 0 0 


 {c} 



   
2
   
 [K 21 ] [K 22 ] [K 23 ]  [M 22 ]
  {d}  = ω  0 0 {d} (3.37)

  
  
   

[K 31 ] [K 23 ] [K 33 ]  {e}  0 0 [M 33 ]  {e} 
   

43
The elements of the stiffness ([K]) and mass ([M ]) matrices are related to the Ritz approximations used and

some basic material properties of the beam that is considered. The stiffness matrix is symmetric, so there

are only six different equations for those elements. The mass matrix is diagonalized so there are only three

different forms. All of these formulas are given below.

∂φui ∂φuj ∂φu ∂φuj ∂φu ∂φuj

Z  
11
Kij = C11 + C55 i + C66 i dV (3.38)
V ∂x ∂x ∂z ∂z ∂y ∂y

∂φui ∂φvj ∂φu ∂φvj

Z  
12
Kij = C12 + C66 i 21
dV = Kji (3.39)
V ∂x ∂y ∂y ∂x

∂φu ∂φw ∂φu ∂φw

Z  
j j
13
Kij = C13 i + C55 i 31
dV = Kji (3.40)
V ∂x ∂z ∂z ∂x

∂φv ∂φvj ∂φv ∂φvj ∂φv ∂φvj

Z  
22
Kij = C22 i + C44 i + C66 i dV (3.41)
V ∂y ∂y ∂z ∂z ∂x ∂x

∂φvi ∂φw ∂φv ∂φw

Z  
j j
23
Kij = C23 + C44 i 32
dV = Kji (3.42)
V ∂y ∂z ∂z ∂y

∂φw ∂φw ∂φw ∂φw ∂φw ∂φw

Z  
i j j j
33
Kij = C33 + C44 i + C55 i dV (3.43)
V ∂z ∂z ∂y ∂y ∂x ∂x

Z
11
Mij = ρφui φuj dV (3.44)
V

Z
22
Mij = ρφvi φvj dV (3.45)
V

44
 Z
33
Mij = ρφw w
i φj dV (3.46)
V

The general form for the Ritz approximation functions will differ for each of the different support condi-

tions that will be considered. The primary base for all of the function employed is a simple power function.

The most basic form of this is given as

φi (x, y, z) = xj y k z l .

These type of functions provide simplicity when evaluating integrals over a parallelepiped. Different terms

were included to ensure proper support conditions, but power functions were utilized as much as possible.

When any combination of values for j, k, and l is considered these power functions can represent any type

of vibrational mode. If all possible combinations are evaluated, the results will include flexural vibration,

shear vibration, and torsional vibration modes about all three axes. For the purposes of the investigation

performed for this thesis, only the transverse flexural modes were of interest. To simplify and reduce the

calculations required, group theory was utilized. Group theory splits all of the possible Ritz functions into

eight different symmetry groups [45]. Parallelepipeds contain three symmetry planes which intersect each

other at right angles. Using these symmetry planes, the possible displacement patterns can be separated

by considering specific combinations of odd or even behavior for each dimension about these planes. For

example, half of the groups follow the pattern that the u displacement is odd in x, even in y, and even in

z. Therefore only odd integers will be considered for j and only even integers for k and l. The v and w

deformation patterns can be split similarly, and through this certain modes of vibration can be isolated.

Therefore, for the calculations performed for this paper, only two groups were considered which represent

the transverse flexural vibration modes [45]. This isolation of the flexural modes was possible for all three

support conditions which were studied.

The three cases that were considered for the study in this paper are fixed-free, fixed-fixed, and simply-

supported. The Ritz functions that were used for each of these, as well as the frequency equations for the

three one-dimensional beam theories, will be presented in the following sections.

3.3.5 Fixed-Free Beam

Although there are ten different end condition combinations, using symmetric and antisymmetric modes,

these can be reduced to four cases. These are free-free, fixed-free (cantilevered), hinged-hinged (simply

45
supported), and fixed-fixed [39]. An in depth study of when the typical beam theories fail for the free-free

case has already been conducted and will not be considered for this paper. Thus only three different support

combinations need to be analyzed to determine a full picture of when the three one-dimensional theories

are insufficient. The frequency equations that will be used for analysis of the Euler-Bernoulli, Rayleigh, and

Timoshenko models were developed by Han and co-workers [39]. Each of these equations will be reiterated

in this summary for reference.

The first case considered will be the cantilevered beam. When the boundary condition discussed in the

previous section were substituted into the spatial solutions, the frequency equations for the free-free case

were found for each of the three one-dimensional theories.

Euler-Bernoulli

For the Euler-Bernoulli case the frequency equation is expressed as

cos a cosh a − 1 = 0. (3.47)

For this beam model, because the relationship between the dimensionless wave number and the frequency is

so simple, the frequency can be found directly using the following equation [39].

s
E∗I ∗ 2
ω∗ = a (3.48)
ρ∗ A∗ L∗4

The ∗ s in this equation signify that these parameters are no longer dimensionless. The dimensionless wave

numbers for the first five modes of vibration were found and are listed below [39].

Table 3.1: Euler-Bernoulli Wave Numbers for Fixed-Free Beam

a1 a2 a3 a4 a5
1.875 4.694 7.855 10.996 14.137

These values were used to find the first five frequencies of vibration for beams with varying slenderness ratios,

which are compared with the values from the other beam models.

46
Rayleigh

The frequency equation for a cantilevered beam using the Rayleigh beam model is given by [39]

(b6 − a6 ) sin a sinh b + 2a3 b3 cos a cosh b − 2a3 b3 = 0. (3.49)

Because this model involves two wave numbers, the process to find the frequency values is not as simple.

The equation for the natural frequency in terms of these two wave numbers is expressed as

a2 − b2
ω2 = . (3.50)
ρI

To simplify the calculation, the wave number b is expressed in terms of a by the relation

r
1
b=a . (3.51)
a2 k 2 + 1

where k is equal to 1/s and s is the slenderness ratio which is given by [39]

r
∗ A∗
s=L . (3.52)
I∗

There are now two equations with three unknowns, so an iterative process is utilized to converge to the

accurate wave numbers and natural frequencies. This process has been performed for multiple beams of

varying slenderness ratio. The first five frequencies are then compared to the other considered beam models.

Timoshenko

Two frequency equations are necessary for the Timoshenko beam model due to the two different frequency

cases discussed in the previous section. The frequency equation for a fixed-free beam when ω < ωc is given

by
a4 + a4 γ 4 + 4γ 2 a2 b2 + b4 γ 4 + b4
(a2 − b2 ) sin a sinh b − ab cos a cosh b − 2ab = 0. (3.53)
(b2 + γ 2 a2 )(a2 + γ 2 b2 )

The equation for ω > ωc is slightly different and is expressed by [39]

a4 + a4 γ 4 − 4γ 2 a2 b̄2 + b̄4 γ 4 + b̄4

(a2 + b̄2 ) sin a sin b̄ − ab̄ cos a cos b̄ − 2ab̄ = 0. (3.54)
(−b̄2 + γ 2 a2 )(a2 − γ 2 b̄2 )

47
Both of these equations use the term γ to simplify the equation. This term is given as

2(1 + ν)
γ2 = . (3.55)
k0

This beam model has two dimensionless wave numbers, similar to the Rayleigh theory, and the problem is

again simplified by finding a relationship between these two values [39]. When ω < ωc this relation is

(γ 2 b2 + a2 )(a2 γ 2 + b2 )
= s2 . (3.56)
(a2 − b2 )(1 + γ 2 )

For the case where ω > ωc the relationship is expressed as

(−γ 2 b̄2 + a2 )(a2 γ 2 − b̄2 )

= s2 . (3.57)
(a2 + b̄2 )(1 + γ 2 )

Using these equations along with the frequency equations, an expression for the natural frequency of a beam

can be found. This expression for the case where the natural frequency is less than the critical frequency is

given by s s
∗ 1 a2 − b2 E∗
ω = ∗ . (3.58)
L 1 + γ2 ρ∗

If the natural frequency is greater than the critical frequency, the express for the natural frequency is [39]

s s
1 a2 + b̄2 E∗
ω∗ = ∗ . (3.59)
L 1 + γ2 ρ∗

With these frequency expressions and the relationships between wave numbers, there are again two equations

and three unknowns for each of the two cases. An iterative process is utilized once more to converge to the

correct answer through these sets of equations. By this procedure, the first five natural frequencies are found

for this theory and are compared with the results from the other three beam models.

Linear Elasticity

As discussed in the previous section, the natural frequencies and modes shapes for the linear elasticity

model will be found by solving a general eigenvalue problem. Suitable Ritz-approximations are necessary

to determine the matrix values for this eigenvalue problem. For the case of a cantilever beam the following

48
general form for the Ritz equations is used.

φi = xj y k z l l>0 (3.60)

For this beam case, the same approximation can be utilized for u, v, and w since the boundary constraints

are identical for all three direction components. For this support case, the number of Ritz terms used to

express each of the displacement components is set by the parameter j + k + l = 16, which provides very

accurate frequency results.

3.3.6 Fixed-Fixed Beam

Using the boundary conditions discussed for a fixed support at both ends of a beam, the frequency

equations have been determined for each of the three typical beam theories. A different form of Ritz-based

approximations has also been found for the application of a fixed-fixed beam. These equations will be

presented in the following paragraphs.

Euler-Bernoulli

The frequency equation for a fixed-fixed beam using the Euler-Bernoulli model is given by

cos a cosh a − 1 = 0. (3.61)

The equation for the natural frequency in terms of the dimensionless wave number a is identical to that used

for the cantilever beam. The first five wave numbers for this support case have been developed by Han and

co-workers and are given below [39].

Table 3.2: Euler-Bernoulli Wave Numbers for Fixed-Fixed Beam

a1 a2 a3 a4 a5
4.730 7.853 10.996 14.137 17.279

The first five natural frequencies using these values have been calculated and are compared to the other

three discussed beam models.

49
Rayleigh

The frequency equation using fixed boundary conditions at both ends of a beam for the Rayleigh model

is expressed as [39]

(b6 − a6 ) sin a sinh b + 2a3 b3 cos a cosh b − 2a3 b3 = 0. (3.62)

The process of determining the natural frequencies and the equations used for this procedure are the same

as discussed in the fixed-free beam case. Using an iterative process the dimensionless wave numbers and

frequencies for the first five modes of vibration have been determined and compared with the other beam

theories.

Timoshenko

For the case of ω < ωc , the frequency equation using the Timoshenko beam model is expressed as

(a2 − b2 )(γ 2 a2 + γ 2 b2 + γ 2 ab − ab)(γ 2 a2 + γ 2 b2 − γ 2 ab + ab)

sin a sinh b − cos a cosh b + 1 = 0. (3.63)
2ab(b2 + γ 2 a2 )(a2 + γ 2 b2

If ω > ωc , the frequency equation becomes

(a2 − b̄2 )[(γ 2 a2 − γ 2 b̄2 )2 + (γ 2 ab̄ − ab̄)2 ]

sin a sin b̄ − cos a cos b̄ + 1 = 0. (3.64)
2ab̄(−b̄2 + γ 2 a2 )(a2 − γ 2 b̄2 )

The expression for the natural frequency is identical to that presented for the fixed-free beam case [39]. Using

this equation and the frequency equation, the first five natural frequencies were found using an iterative

process similar to that discussed for the cantilever beam.

Linear Elasticity

The general form of the Ritz approximation terms is similar to that of the fixed-free case, but it has

been altered to fulfill the proper fixed end boundary conditions at the second end. The origin of the axes

is again set at one end of the beam, so the approximation terms for this support case need to provide zero

displacement and slope at z = 0 and z = L. The general form that was selected to allow for this is given as

φi = xj y k z l (z − L)2 . (3.65)

This approximation form can again be used for all three direction components, u, v, and w for the fixed-

50
fixed beam case since the boundary conditions are the same for all three directions in a fixed end. The same

parameter of j + k + l = 16 that was used to determine the number of terms in the approximation for the

cantilever beam is again used for the fixed-fixed beam. This provides for very accurate frequency and mode

shape solutions when the general eigenvalue problem is solved for the linear elasticity beam model. These

solutions will be compared to the typical theories to understand how the assumptions in those theories affect

their accuracy.

3.3.7 Simply-Supported Beam

For the simply-supported beam case, the boundary conditions of no transverse displacement and no

moment are applied at both ends of the beam. Substituting these requirements into the spatial solutions

allows for the determination of the frequency equation for each beam model. Ritz approximation forms for

the linear elasticity model of a simply-supported beam are also discussed.

Euler-Bernoulli

The frequency equation for a simply-supported beam as developed by Han and co-workers is given by

[39]

sin a sinh a = 0. (3.66)

The equation for the natural frequency from the previous two cases can also be used for this support condition.

The first five dimensionless wave numbers that can be substituted into this equation are listed below [39].

Table 3.3: Euler-Bernoulli Wave Numbers for Simply-Supported Beam

a1 a2 a3 a4 a5
π 2π 3π 4π 5π

Rayleigh

For the Rayleigh beam model, the simply-supported frequency equation is expressed as [39]

sin a sinh b = 0. (3.67)

Using the relation between wave numbers and the equation for natural frequency presented before, the first

51
five natural frequencies have been found through iteration to converge on the solution. These values are

compared with the other beam models for the simply-supported case.

Timoshenko

When the natural frequency is less than the critical value, the frequency equation for a simply-supported

beam using the Timoshenko theory is identical to the Rayleigh model equation as seen below.

sin a sinh b = 0 (3.68)

For the case when the frequency is greater than ωc the frequency equation is given by [39]

sin a sin b̄ = 0. (3.69)

With the natural frequency equation that was presented for the two cases in the Timoshenko model, the

above frequency equations can be used to determine the first five modes of vibration. An iterative procedure

is again utilized to converge to the solutions for the simply-supported case. These solutions are compared

with the other models, especially the linear elasticity case, to study the accuracy of this beam model.

Linear Elasticity

The Ritz-based approximations for a simply-supported beam are significantly different than for the other

two support cases considered, because the constraints are not the same for all three direction components

for a hinged end. Therefore, the same general form of approximation terms cannot be employed for all three

directions. Another considerable difference for the simply-supported case comes from the inability to satisfy

the necessary boundary conditions through a power series. In order to satisfy these conditions and still have

the ability to simply evaluate the functions over a parallelepiped, trigonometric functions are utilized for

the axial dimension. Sine functions are used for the transverse and out-of-plane displacements, and cosine

terms will be employed for the axial displacement. The type of trigonometric function that is used for

each displacement component was determined by studying the displacement patterns from a finite element

analysis of simply-supported beam. With all of these considerations, the form of the approximations are as

follows.

 
lπz
φui j k
= x y sin (3.70)
L

52
 

lπz
φvi = xj y k sin (3.71)
L
 
lπz
φw
i = xj k
y cos (3.72)
L

A slightly different parameter of j +k +l = 14 was used to set the number of terms used in the approximation

for each displacement coordinate. This was changed because using a limit of j + k + l = 16 resulted in so

many terms in each approximation that there was not enough memory on the server used for calculation to

perform the analysis. The results of this linear elasticity theory are compared to the typical one-dimensional

theories, particularly for less slender beams to study the accuracy of both the natural frequencies and modes

shapes of vibration.

3.4 Conclusion

Four different beam theories have been presented and discussed in this chapter. Using the information pre-

sented in this chapter, a study has been conducted to investigate when three typically used, one-dimensional

beam theories begin to deteriorate in accuracy. To complete this study, the three beam theories have been

compared with a fully three-dimensional, linear elasticity beam model that utilizes Ritz-based approxima-

tions to calculate highly accurate natural frequencies and mode shapes for the first five modes of vibration

of beams with varying dimensions. The results and conclusions from this vibration study are discussed in

Chapter 4.

53
Chapter 4

Methods and Results

4.1 Introduction

A study of vibrational mechanic theories was performed to investigate their accuracy and effectiveness.

This analysis was performed for three different support cases: fixed-free, fixed-fixed, and simply-supported.

For each of these cases the frequencies of the first five flexural vibration modes were computed using the

Euler-Bernoulli, Rayleigh, and Timoshenko beam theories, as well as a full elasticity approximation with and

without the inclusion of the Poisson effect. One of the assumptions utilized in the three common theories

that may cause inaccuracies is that a beam is slender. One purpose of this study was to investigate when

this assumption leads to significant errors in the frequency results. Therefore, the modal frequencies were

found for a variety of slenderness ratios, ranging from less than 7 to almost 350. Another item of interest in

this study was how the assumption of an isotropic material affects the accuracy of typical beam theories. To

investigate this the resonant frequencies were found for both and isotropic material (steel) and an orthotropic

material (graphite-magnesium). The last assumption which was examined in this study was that the Poisson

effect was negligible. This was studied by utilizing the Ritz-based approximation with a zero Poisson ratio

as well as the typical material Poisson ratios. The results of these investigations are presented and discussed

in the remainder of this chapter.

Also examined in this chapter are the results of calculations performed for a potential new infrastructure

sensor. The device is a microelectromechanical system (MEMS) sensor, which will operate by exciting

vibrational resonance in a small fixed-fixed bridge within the device using a microwave-frequency current.

In order to properly excite the bridge, accurate resonant frequencies are needed. Using the full elasticity

approximation presented in the vibrational mechanics study, the resonant frequencies for flexure in the lateral

modes have been determined. This information will be useful in future testing and calibration of the device.

54
4.2 Vibration Study Methods

As mentioned in the previous section, the accuracy of typical beam theories was investigated by deter-

mining the resonant frequencies of the first five modes for varying beam types using each theory. In order to

calculate these results, computing programs were prepared for each theory and support case based upon the

governing equations presented in chapter three. Each of these programs were written to allow for different

beam geometries and material properties.

The assumption of one beam dimension being significantly larger than the other two was the first to be

investigated. This was done by considering beams of many different lengths, and thus varying slenderness

ratios. Each calculation was done for a one by one cross section. The different lengths that were studied

include: 2, 5, 10, 20, 40, and 100. Calculations using the three typical theories, as well as a full elasticity

approximation were performed for each of these lengths and compared. These results are discussed in the

next section of this report.

The three beam theories of interest in this study also assume that the material of a beam is isotropic.

An anisotropic material can be somewhat accommodated through the shear coefficient for the Timoshenko

model, but all other beam theory equations do not consider the effect of anisotropy. This vibrational study

therefore, also considered the accuracy of these models when a beam is both isotropic or orthotropic. Steel

was chosen as the isotropic material to study, while graphite-magnesium was utilized for an orthotropic

material. The elastic constants and stiffness values for a graphite-magnesium composite were determined by

H.M. Ledbetter and others [46]. These values, as well as those for steel, are given in Table 4.1.

The last item of particular interest in this study is the Poisson effect. The three theories typically used

for beam analysis neglect this effect. The full elasticity approximation allows for displacements due to

the Poisson effect and thus allowed for an investigation of what involvement this phenomenon has on the

frequency and mode shapes of a vibrating beam. Also, to attempt to isolate the influence of the Poisson

effect from the other two assumptions previously discussed, the Ritz-based elasticity approximation was used

to calculate frequencies with zero ν as well as the with the correct Poisson ratios listed in Table 4.1. The

findings of this study are presented in the following section.

4.3 Vibration Study Results

The resonant frequencies for the first five modes were the primary results that were investigated in this

vibrational mechanics study. These values were found for various lengths, materials, and support conditions.

55
Table 4.1: Material Properties for Vibration Study, units are GPa except for dimensionless νij and Density
in kg/m3

Constant Steel Graphite-Magnesium

E1 200 23.81
E2 200 23.81
E3 200 166.64
ν12 0.3 0.359
ν13 0.3 0.045
ν23 0.3 0.045
ν21 0.3 0.359
ν31 0.3 0.314
ν32 0.3 0.314
C11 269.23 28.19
C22 269.23 28.19
C33 269.23 174.30
C44 76.923 17.91
C55 76.923 17.91
C66 76.923 8.76
C12 115.38 10.67
C13 115.38 12.20
C23 115.38 12.20
Density, ρ 7830 1738

All of the calculated frequencies can be found in the following tables. The results for a cantilever beam

are listed in Table 4.2 and Table 4.3, for an isotropic and orthotropic beam, respectively. For a fixed-fixed

support case, the results are provided in Table 4.4 and Table 4.5, again for isotropic or orthotropic materials.

Lastly, the frequencies for a simply-supported beam case are given in Table 4.6 and Table 4.7. For most of

the cases involving a fixed support, the higher mode frequencies from the Timoshenko beam model were not

able to be calculated. It seems that when the slenderness ratio becomes so small, the program created to

determined the Timoshenko values will not converge to an answer. Even without some of these values trends

were still able to be determined and conclusions drawn. From the values found the influence of slenderness,

anistropy, and the Poisson effect could be investigated. The effect of each of these factors for the different

support conditions are discussed in the remainder of this section.

56
 Table 4.2: Frequencies of an Isotropic Cantilevered Beam
L Mode E-B Rayleigh Timoshenko ν=0 Full
1 1282.29 1224.17 1093.04 1099.82 1149.29
2 8036.55 6205.44 4147.35 4284.97 4313.85
2 3 22501.84 14104.63 8676.21 8898.61 8919.19
(S=6.928) 4 44101.46 22570.04 11251.97 11306.62 11320.33
5 72895.05 31184.49 14596.18 13161.22 13396.42
1 205.19 203.62 199.08 199.33 209.11
2 1285.87 1221.42 1074.13 1089.48 1121.12
5 3 3600.77 3211.34 2560.16 2634.54 2681.66
(S=17.32) 4 7056.23 5812.27 4255.09 4446.58 4501.16
5 11663.21 8811.58 6055.60 6276.84 6312.54
1 51.297 51.198 50.900 50.924 53.452
2 321.467 317.215 305.250 307.096 318.146
10 3 900.193 872.477 803.480 815.589 836.107
(S=34.64) 4 1764.058 1667.443 1458.715 1501.403 1526.117
5 2915.805 2673.048 2222.277 2382.836 2407.668
1 12.824 12.818 12.799 12.803 13.440
2 80.367 80.099 79.291 79.570 82.617
20 3 225.048 223.244 218.069 220.133 226.432
(S=69.28) 4 441.015 434.556 416.846 425.520 433.694
5 728.951 712.239 668.673 712.071 719.086
1 3.2061 3.2057 3.2045 3.2053 3.3647
2 20.0917 20.0752 20.0237 20.0807 20.8626
40 3 56.2621 56.1457 55.8054 56.2424 57.9127
(S=138.56) 4 110.2536 109.8364 108.6221 110.5882 112.8251
5 182.2378 181.1688 178.0269 189.1699 190.9816
1 0.51297 0.51296 0.51293 0.51305 0.53857
2 3.21467 3.21431 3.21298 3.22153 3.34755
100 3 9.00193 8.99848 8.98964 9.05576 9.32763
(S=346.41) 4 17.64058 17.62862 17.59666 17.90082 18.26871
5 29.15805 29.13104 29.04698 30.84565 31.13821

57
 Table 4.3: Frequencies of an Orthotropic Cantilevered Beam
L Mode E-B Rayleigh Timoshenko ν=0 Full
1 2484.37 2371.78 1703.94 1729.30 1738.72
2 15570.42 12022.72 5434.23 5548.43 5556.60
2 3 43602.01 27327.02 - 9726.02 9801.51
(S=6.928) 4 85444.40 43728.35 - 10818.68 10871.55
5 141230.55 60418.40 - 11667.13 11715.13
1 397.50 394.50 365.77 366.76 369.45
2 2491.27 2366.44 1659.44 1707.68 1713.45
5 3 6976.32 6221.81 3603.82 3708.18 3715.05
(S=17.32) 4 13671.10 11260.99 5570.92 5677.38 5688.80
5 22596.89 17072.00 7556.90 7700.27 7714.03
1 99.375 99.194 97.203 97.281 98.043
2 622.817 614.588 542.805 549.257 551.907
10 3 1744.080 1690.380 1324.954 1359.684 1364.268
(S=34.64) 4 3417.776 3230.588 2240.531 2335.581 2342.316
5 5649.222 5178.898 3220.536 3337.431 3343.200
1 24.844 24.834 24.707 24.715 24.912
2 155.704 155.187 148.884 150.658 151.499
20 3 436.020 432.524 400.108 405.525 407.119
(S=69.28) 4 854.444 841.931 737.033 757.386 759.503
5 1412.305 1379.927 1136.833 1219.862 1222.479
1 6.2109 6.2109 6.2028 6.2044 6.2540
2 38.9260 38.8947 38.5475 38.6736 38.8988
40 3 109.0050 108.7795 106.5160 107.4695 107.9194
(S=138.56) 4 213.6110 212.8027 204.8697 208.9977 209.5795
5 353.0764 351.0056 330.9243 352.5038 353.0551
1 0.99375 0.99384 0.99363 0.99385 1.00182
2 6.22817 6.22757 6.21856 6.23551 6.27229
100 3 17.44080 17.43411 17.37419 17.50519 17.57998
(S=346.41) 4 34.17776 34.15457 33.93870 34.53706 34.63319
5 56.49222 56.43994 55.87438 59.35934 59.43626

58
 Table 4.4: Frequencies of an Isotropic Fixed-Fixed Beam
L Mode E-B Rayleigh Timoshenko ν=0 Full
1 8160.29 7276.64 3947.00 4067.61 4156.69
2 22493.38 16031.52 7711.15 7951.97 8016.46
2 3 44101.46 25232.54 - 12665.35 12740.30
(S=6.928) 4 72895.05 34162.23 - 17869.87 18071.29
5 108898.19 42894.24 - 23039.24 23231.10
1 1305.65 1279.67 1053.57 1070.31 1105.72
2 3598.94 3350.56 2415.61 2471.41 2526.54
5 3 7056.23 6117.99 4029.29 4157.36 4224.59
(S=17.32) 4 11663.21 9302.02 5760.38 5961.74 6038.66
5 17423.71 12708.68 7566.97 7872.99 7936.89
1 326.412 324.756 306.315 310.131 321.455
2 899.735 882.988 785.788 798.092 823.077
10 3 1764.059 1695.419 1421.532 1449.363 1483.411
(S=34.64) 4 2915.802 2727.330 2160.254 2212.983 2254.271
5 4355.928 3943.480 2968.401 3067.753 3117.107
1 81.603 81.500 80.250 81.192 84.220
2 224.934 223.874 216.591 219.501 227.318
20 3 441.015 436.508 413.359 419.157 430.819
(S=69.28) 4 728.950 716.276 661.887 674.313 690.634
5 1088.982 1060.184 954.204 975.394 995.855
1 20.4007 20.3946 20.3147 20.5686 21.3294
2 56.2334 56.1691 55.6898 56.4464 58.5207
40 3 110.2537 109.9629 108.3754 109.8344 113.0188
(S=138.56) 4 182.2376 181.4330 177.5127 180.7184 185.5264
5 272.2455 270.3853 262.3055 267.3909 273.5839
1 3.26412 3.26401 3.26195 3.32304 3.43547
2 8.99735 8.99609 8.98364 9.14956 9.47477
100 3 17.64059 17.63206 17.59028 17.96247 18.44485
(S=346.41) 4 29.15802 29.13788 29.03299 29.77728 30.58208
5 43.55928 43.51023 43.28975 44.74848 45.75155

59
 Table 4.5: Frequencies of an Orthotropic Fixed-Fixed Beam
L Mode E-B Rayleigh Timoshenko ν=0 Full
1 15810.17 14098.13 4500.41 4649.87 4661.04
2 43579.82 31060.28 8885.38 9192.75 9211.47
2 3 85444.40 48886.80 - 10574.52 10683.07
(S=6.928) 4 141230.55 66187.62 - 17038.85 16722.03
5 210984.86 83105.47 - 21824.63 21772.46
1 2529.63 2479.30 1508.23 1537.40 1544.50
2 6972.77 6491.54 3135.55 3222.04 3230.33
5 3 13671.10 11853.31 5014.83 5142.33 5155.16
(S=17.32) 4 22596.89 18022.20 6937.85 7135.95 7150.96
5 33757.58 24622.44 8885.12 9138.90 9155.73
1 632.407 629.200 524.644 531.315 534.244
2 1743.192 1710.746 1230.568 1253.115 1258.302
10 3 3417.776 3284.789 2080.759 2135.417 2142.921
(S=34.64) 4 5649.222 5284.067 3002.296 3089.447 3099.222
5 8439.394 7640.297 3963.226 4120.143 4129.059
1 158.102 157.902 149.847 151.556 152.440
2 435.798 433.745 389.357 394.379 396.684
20 3 854.444 845.713 713.313 725.179 728.129
(S=69.28) 4 1412.305 1387.750 1096.113 1119.157 1123.014
5 2109.849 2054.054 1519.743 1564.651 1569.730
1 39.5254 39.5134 38.9782 39.4260 39.6587
2 108.9495 108.8248 105.6627 107.0223 107.6105
40 3 213.6110 213.0477 202.7912 205.4340 206.3256
(S=138.56) 4 353.0764 351.5174 326.8051 332.5416 333.7862
5 527.4621 523.8582 474.2981 484.0115 485.6946
1 6.32407 6.32385 6.30990 6.39440 6.43117
2 17.43192 17.42949 17.34520 17.59123 17.68912
100 3 34.17776 34.16125 33.87952 34.36394 34.51226
(S=346.41) 4 56.49222 56.45319 55.74922 56.79759 57.01745
5 84.39394 84.29893 82.82703 84.53273 84.82903

60
 Table 4.6: Frequencies of an Isotropic Simply-Supported Beam
L Mode E-B Rayleigh Timoshenko ν=0 Full
1 3599.84 3278.53 2712.61 2705.03 2688.52
2 14399.36 10666.29 7393.83 7367.28 7286.58
2 3 32398.56 19189.40 12229.28 12199.06 12060.85
(S=6.928) 4 57597.45 27808.74 17010.07 16997.14 16831.98
5 89996.01 36318.21 21730.55 21749.33 21584.08
1 575.97 566.73 541.69 541.23 540.43
2 2303.90 2165.80 1881.89 1877.67 1869.13
5 3 5183.77 4553.32 3600.91 3589.39 3562.46
(S=17.32) 4 9215.59 7459.20 5470.54 5450.81 5398.13
5 14399.36 10666.29 7393.83 7367.28 7286.64
1 143.994 143.405 141.662 141.629 141.572
2 575.974 566.727 541.687 541.232 540.432
10 3 1295.943 1250.487 1142.155 1140.365 1137.004
(S=34.64) 4 2303.898 2165.797 1881.892 1877.666 1869.127
5 3599.840 3278.826 2712.610 2705.026 2688.526
1 35.998 35.961 35.849 35.847 35.843
2 143.994 143.405 141.662 141.629 141.572
20 3 323.986 321.029 312.605 312.446 312.174
(S=69.28) 4 575.974 566.728 541.687 541.232 540.433
5 899.960 877.684 820.892 819.907 818.118
1 8.9996 8.9973 8.9902 8.9901 8.9899
2 35.9984 35.9615 35.8494 35.8487 35.8434
40 3 80.9964 80.8097 80.2490 80.2380 80.2196
(S=138.56) 4 143.9936 143.4051 141.6625 141.6285 141.5716
5 224.9900 223.5581 219.3924 219.3123 219.1770
1 1.43994 1.43987 1.43970 1.43969 1.43969
2 5.75974 5.75880 5.75590 5.75585 5.75575
100 3 12.95943 12.95463 12.94001 12.93972 12.93924
(S=346.41) 4 23.03898 23.02383 22.97776 22.97684 22.97532
5 35.99840 35.96145 35.84935 35.84712 35.84344

61
 Table 4.7: Frequencies of an Orthotropic Simply-Supported Beam
L Mode E-B Rayleigh Timoshenko ν=0 Full
1 6974.51 6351.98 3854.06 3823.95 3812.57
2 27898.05 20665.41 8856.85 8864.57 8843.49
2 3 62770.61 37178.51 13708.31 10445.80 10573.31
(S=6.928) 4 111592.20 53878.05 18489.92 16619.39 16102.21
5 174362.82 70364.70 21013.71 20934.36 21035.53
1 1115.92 1098.01 950.72 946.36 945.06
2 4463.69 4196.12 2840.57 2817.92 2810.20
5 3 10043.30 8821.84 4869.36 4836.29 4821.75
(S=17.32) 4 17854.75 14451.82 6879.00 6856.22 6837.33
5 27898.05 20665.41 8856.85 8864.57 8843.53
1 278.981 277.840 266.498 266.083 265.964
2 1115.922 1098.007 950.722 946.358 945.059
10 3 2510.825 2422.757 1852.608 1839.732 1835.667
(S=34.64) 4 4463.688 4196.125 2840.571 2817.916 2810.202
5 6974.513 6351.982 3854.063 3823.954 3812.579
1 69.745 69.674 68.920 68.891 68.882
2 278.981 277.840 266.498 266.083 265.964
20 3 627.706 621.978 569.658 567.917 567.410
(S=69.28) 4 1115.922 1098.007 950.722 946.358 945.059
5 1743.628 1700.470 1384.761 1376.555 1374.045
1 17.4363 17.4318 17.3839 17.3820 17.3815
2 69.7451 69.6735 68.9199 68.8906 68.8823
40 3 156.9265 156.5648 152.8476 152.7067 152.6667
(S=138.56) 4 278.9805 277.8403 266.4975 266.0826 265.9642
5 435.9070 433.1328 406.6094 405.6809 405.4136
1 2.78981 2.78969 2.78846 2.78841 2.78839
2 11.15922 11.15739 11.13774 11.13696 11.13673
100 3 25.10825 25.09896 24.99995 24.99604 24.99492
(S=346.41) 4 44.63688 44.60754 44.29649 44.28428 44.28081
5 69.74513 69.67353 68.91992 68.89057 68.88225

62
4.3.1 Effects of Slenderness

From the results of this study it is clear that the slenderness ratio of a beam has a significant effect

on the accuracy of frequency calculations for two of the common beam theories. As the slenderness ratio

decreases to a point where the solid is closer to a cube than a typical beam, the error between the elasticity

solution and both the Euler-Bernoulli and Rayleigh results increases exponentially. For the Timoshenko

beam model, the length of a beam has a minimal effect on the percent error from the full elasticity theory.

This is especially true for the fundamental mode of an isotropic beam. The support case for a beam affects

the level of accuracy for each beam theory as the slenderness ratio changes, but these general trends are

consistent throughout.

For the cantilever beam and fixed-fixed beam scenarios, all three beam theories approach a similar error

value as the slenderness increases. The existence of even one fixed support for a beam leads to noticeable

errors in frequency for even very slender beams. This is most likely due to neglecting the Poisson effect and

will be discussed in Section 4.3.3. Although the frequency results from the beam theories do not approach

the full elasticity solution, they still demonstrate the slenderness trends discussed above. As the slenderness

value decreases, the results for both the Euler-Bernoulli and Rayleigh models begin to deviate from the

asymptotic error percentage at an exponential rate as can be seen in Figures 4.1 and 4.3. The error values

from the Rayleigh beam model are less than those from Euler-Bernoulli model calculations, as was expected,

but the slenderness ratio of a beam still has a meaningful influence on the accuracy of the Rayleigh model.

This effect is significantly greater for higher modes of vibration for both beam theories. While the errors for

the fundamental mode of an 1x1x2 beam are 11.6% and 6.5% for Euler-Bernoulli and Rayleigh, respectively,

these errors grow to 444% and 133% for the fifth mode of vibration in a cantilevered beam. The errors for a

fixed-fixed beam are even larger for higher modes of vibration. The error values for the fifth mode are shown

in Figures 4.2 and 4.4. The minute effect that slenderness has upon the Timoshenko frequency results can

also be seen in these figures. While the other two beam theories diverge exponentially as the beam length

falls below ten, the Timoshenko model remains at a very similar error value throughout the entire range of

slenderness ratios considered. The error increases slightly more for very stout beams and higher modes of

vibration, but this increases is small in comparison to the other two theories.

The results from a simply-supported beam also follow similar trends. The main difference for this support

case is that the frequency values from the three beam theories actually approach the full elasticity solutions,

rather than a common error value. The error still increases significantly for the Euler-Bernoulli and Rayleigh

beam theories as a beam becomes less slender, while the influence is minimal for the Timoshenko model.

63
 12
Euler−Bernoulli
10 Rayleigh

Percent Error from Full Elasticity Solution, %

Timoshenko
8

−2

−4

−6
0 10 20 30 40 50 60 70 80 90 100
Beam Length

Figure 4.1: Frequency Error for the First Mode of a Cantilevered Beam

450
Euler−Bernoulli
400 Rayleigh
Percent Error from Full Elasticity Solution, %

Timoshenko
350

300

250

200

150

100

50

−50
0 10 20 30 40 50 60 70 80 90 100
Beam Length

Figure 4.2: Frequency Error for the Fifth Mode of a Cantilevered Beam

64
 100
Euler−Bernoulli
Rayleigh

Percent Error from Full Elasticity Solution, %

80 Timoshenko

60

40

20

−20
0 10 20 30 40 50 60 70 80 90 100
Beam Length

Figure 4.3: Frequency Error for the First Mode of a Fixed-Fixed Beam

400
Euler−Bernoulli
350 Rayleigh
Percent Error from Full Elasticity Soltuion, %

Timoshenko
300

250

200

150

100

50

−50
0 10 20 30 40 50 60 70 80 90 100
Beam Length

Figure 4.4: Frequency Error for the Fifth Mode of a Fixed-Fixed Beam

65
 35
Euler−Bernoulli
Rayleigh

Percent Error from Full Elasticity Solution, %

30
Timoshenko

25

20

15

10

0
0 10 20 30 40 50 60 70 80 90 100
Beam Length

Figure 4.5: Frequency Error for the First Mode of a Simply-Supported Beam

Slenderness is also still a greater factor for higher modes of vibration than for the fundamental mode. The

error values for a simply-supported beam as the beam length varies are shown in Figures 4.5 and 4.6.

All of these results follow the expected trends. As the slenderness of a beam decreases, the displacement

pattern is known to deviate from the linear model that each of the three beam theories assume. For

example, in the extreme case where the beam length approaches the width or becomes less than the width,

the deformation pattern becomes more similar to that of a plate than a beam. It is therefore clear that the

one-dimensional displacement patterns of the common beam theories cannot be applied to very short and

thick beams. In this study, three-dimensional displacement patterns were determined using the Ritz-based

approximation program. It was found that obvious warping of the beam surface through the width and height

begins to occur for very stocky beams. When a beam is very slender, the displacement is dominant in only

one dimension as the three common beam theories assume, but this deteriorates rapidly as the slenderness

decreases. The first five three-dimensional mode shapes were plotted for both a very slender beam (1x1x40)

and a very stout beam (1x1x2) for each of the support cases. Figures 4.7, 4.8, 4.9, 4.10, 4.11, and 4.12

show these mode shapes and how the deformation behavior drastically changes for non-slender beams. You

can clearly see how the deformation pattern deviates from beam theory expectations in Figure 4.8 where

the cross-sectional plane is no longer plane throughout displacement; this is especially noticeable on the

66
 350
Euler−Bernoulli
Rayleigh

Percent Error from Full Elasticity Soltuion, %

300
Timoshenko

250

200

150

100

50

0
0 10 20 30 40 50 60 70 80 90 100
Beam Length

Figure 4.6: Frequency Error for the Fifth Mode of a Simply-Supported Beam

free face of the fifth mode shape. In the fourth and fifth mode shape for both the fixed-fixed and simply-

supported cases, the Poisson effect begins to have a significant influence for very short beams. In these plots,

obvious warping through the width of the beam occurs due to the shrinking and expansion influences of

the Poisson ratio. The displacement patterns in these cases can no longer be accurately modeled through

a one-dimensional plot of the centerline since the deformation of the outer surfaces is significantly different

than what occurs at the centerline of the beam. With the very slender beams, the deformation follows

typical beam theory expectations with the entire cross section of the beam moving essentially as one. In

very slender beam cases, it is therefore acceptable to use a one dimensional centerline displacement to model

beam vibrations. Three dimensional mode shapes are necessary though for low slenderness ratios to capture

the accurate displacements.

An unexpected result did arise when creating the three-dimensional modal deformation plots. For most

beams, the lower modes of vibrational are always flexural, but it was found that for very non-slender beams

this was not the case. When determining the mode shapes for the 1x1x40 beam, the lowest five modal

frequencies were clearly flexural modes and matched exactly the expected displacement patterns. For the

very short beam case, the lowest three modes followed flexural displacements for all three support conditions,

but the fourth and fifth lowest frequencies corresponded to very different deformation types. For the fixed-

67
fixed case, the fifth flexural mode did not appear until the 14th lowest mode of vibration. This finding

provides another strong reason for more accurate frequency and mode shape calculations when dealing with

non-slender beams since non-flexural modes of vibration may become dominant far earlier than anticipated.

If the only factor affecting the accuracy of frequency results were slenderness, it would be clear that the

Timoshenko model provides an acceptable approximation for frequencies. This model presents a significant

improvement over either of the other two beam models when it comes to the loss of accuracy with less slender

beams. The Rayleigh model gives improved results in comparison to Euler-Bernoulli, but the correction is

minor compared to that of the Timoshenko model. The slenderness ratio of a beam is not the only element

affecting the frequency results, though, so the conclusion that the Timoshenko model is adequate cannot

be made. The material properties and the Poisson effect also alter the results as will be discussed in the

following paragraphs.

4.3.2 Effects of Anisotropy

The slenderness ratio of a beam has arguably the greatest influence on the accuracy of frequency results,

but the results of this study demonstrate that anisotropic material properties can intensify this effect. Since

the three basic theories assume that a beam is composed of isotropic material, the error in frequency results

increases significantly when they are applied to an orthotropic material. This is especially true for the

Euler-Bernoulli and Rayleigh beam theories. Figures 4.13, 4.14, and 4.15 display this trend. The accuracy

of frequency values for the orthotropic material, graphite-magnesium, decreases even more rapidly than for

steel as the slenderness decreases. For example, the error is more than four times greater for the orthotropic

fixed-fixed beam than the isotropic case. The error increase for an anisotropic beam becomes insignificant

for very slender beams, displaying that the slenderness ratio provides the dominant influence on frequency

results. Anisotropic material properties simply exacerbate the loss of accuracy at low slenderness ratios.

A result from this study which was unexpected, comes from the Timoshenko beam model. Surprisingly,

the Timoshenko model was exceptionally adaptable to the orthotropic material used. It is unclear whether

this outcome is unique to graphite-magnesium or if the Timoshenko beam theory would perform well for any

orthotropic or anisotropic material. More calculations would need to be completed with a larger variety of

materials to answer that question. For the materials considered in this study, though, the frequency errors

were very similar for both the isotropic and orthotropic material as can be noted from Figures 4.13, 4.14,

and 4.15. The fact that the Timoshenko model could accommodate an orthotropic material may be due to

the influence of the shear factor. As discussed in chapter 3, the shear coefficient for the orthotropic material

68
Figure 4.7: First Five Mode Shapes for a 1x1x40 Figure 4.8: First Five Mode Shapes for a 1x1x2
Isotropic Fixed-Free Beam Isotropic Fixed-Free Beam

69
Figure 4.9: First Five Mode Shapes for a 1x1x40 Figure 4.10: First Five Mode Shapes for a 1x1x2
Isotropic Fixed-Fixed Beam Isotropic Fixed-Fixed Beam

70
Figure 4.11: First Five Mode Shapes for a 1x1x40 Figure 4.12: First Five Mode Shapes for a 1x1x2
Isotropic Simply-Supported Beam Isotropic Simply-Supported Beam

71
 45
E−B Iso
40 E−B Ortho

Percent Error from Full Elasticity Solution, %

Rayleigh Iso
35 Rayleigh Ortho
Timo Iso
30 Timo Ortho

25

20

15

10

−5
0 10 20 30 40 50 60 70 80 90 100
Beam Length

Figure 4.13: Error Comparison Between Isotropic and Orthotropic Materials for a Cantilevered Beam

250
E−B Iso
E−B Ortho
Percent Error from Full Elasticity Solution, %

200 Rayleigh Iso

Rayleigh Ortho
Timo Iso
150 Timo Ortho

100

50

−50
0 10 20 30 40 50 60 70 80 90 100
Beam Length

Figure 4.14: Error Comparison Between Isotropic and Orthotropic Materials for a Fixed-Fixed Beam

72
 250
E−B Iso
E−B Ortho

Percent Error from Full Elasticity Solution, %

200 Rayleigh Iso
Rayleigh Ortho
Timo Iso
150 Timo Ortho

100

50

−50
0 10 20 30 40 50 60 70 80 90 100
Beam Length

Figure 4.15: Error Comparison Between Isotropic and Orthotropic Materials for a Simply-Supported Beam

was calculated through a number of equations that took the anisotropic material properties into account.

Neither of the other two beam theories provide a means to consider anything other than isotropic properties,

which may be the reason for the significant loss of accuracy when applied to an orthotropic beam.

The error increase for both the Euler-Bernoulli and Rayleigh beam theories is consistent throughout

the different support conditions investigated for this study. The percent by which the frequency error

increases varies slightly for the three cases, but in general the trend is consistent. For the Euler-Bernoulli

and Rayleigh beam theories, the effect of an anisotropic material significantly amplifies the influence of low

slenderness upon the accuracy of frequency results. For the Timoshenko model, the shear coefficient allows

for the inclusion of anisotropic properties and thus provides similar results to isotropic beams for all three

support cases. If the impacts of slenderness and anisotropy were the only considerations, it seems as if the

Timoshenko model would be a reasonable approximation of frequency results for even very stout orthotropic

beams. These two effects cannot be isolated though, and all three models result in significant error values

for all slenderness values if at least one of the beam supports is fixed. The primary cause for this accuracy

problem is most likely due to neglecting the Poisson effect, which will be discussed in the next section.

73
4.3.3 Effects of Poisson Ratio

The most unexpected result from this vibrational mechanics study is the significant difference between

the Ritz-based elasticity method and the three beam theories for even very slender beams when at least one

support in a beam is fixed. When these large errors were discovered, the suggested cause was the assumption

of zero Poisson ratio in the typical beam theories. When considering a three-dimensional model, a perfectly

fixed support completely restricts the Poisson effect. For example, in a cantilevered beam, the largest forces

would exist at the fixed support and thus the influence of the Poisson effect would cause the beam to shrink

or expand through the width to counteract the large axial compression and tension forces. A perfectly fixed

end does not allow this though, so the displacements and forces cannot act as is expected by typical beam

theories. With a hinged or free support, there is no restriction of the Poisson effect and the beam is free

to displace and vibrate as expected without any unusual effects. Therefore, ignoring the Poisson effect in a

beam with only free or hinged supports would have minimal effect on the accuracy of the frequency results.

In order to investigate this theory, the more accurate Ritz-based elasticity program was used to determine

frequency results with both the normal material properties listed in Table 4.1 as well as elastic stiffness

values found using ν = 0. This allowed for the influence of the Poisson effect to be isolated from the other

factors that cause errors in the typical beam theories.

The results of this investigation show that a consistent error exists between the models with and without

the Poisson effect that is independent of beam slenderness. This is displayed in Figure 4.16, where the error

in the fundamental mode is near 4.7% for a cantilevered beam, 3.4% for a fixed-fixed beam, and 0% for

a simply-supported beam. The percent difference does vary somewhat with the length of the beam, but

the change is minimal and does not seem to follow a persistent trend. The result for the simply-supported

beam case followed expectations and showed that neglecting the Poisson effect has a minimal influence. This

was also the result for a free-free beam which was investigated prior to this study. The error results were

somewhat surprising for the fixed-free and fixed-fixed support conditions. Both of these cases did show a

considerable error for all slenderness values, which was predicted, but the expected result was that the error

would be higher for the fixed-fixed case than for the cantilevered beam. It seemed that the influence of the

Poisson effect would increase with more restriction at the supports, but this was not the case. The reason

for this is unclear and more study would be required to understand why the error is greater for the fixed-free

case than for a fixed-fixed beam.

Although, the impact of neglecting the Poisson effect seems to be independent of slenderness, the error is

related to the mode of vibration. Unexpectedly, the error decreased for higher modes of vibration for both

74
 5

Percent Error for Zero Poisson Ratio, %

3

2 Fixed−Free
Fixed−Fixed
Simply−Supported
1

−1
0 10 20 30 40 50 60 70 80 90 100
Beam Length

Figure 4.16: Influence of Poisson Effect for Varying Slenderness Ratios

cases including a fixed support. It is uncertain why this is the case, and more investigation is necessary to

determine the reason behind this pattern. For the simply-supported case, the opposite pattern occurs. The

influence of the Poisson effect is still minor for all modes of vibration, but the error does tend to increase for

higher modes of vibration. These results are shown for a beam with a slenderness ratio of 69 in Figure 4.17.

The frequency error due to the Poisson ratio also appears to be connected with the material properties

of a beam when a fixed support is present. The percent difference between the zero Poisson ratio and

full elasticity results is far less for the orthotropic material studied than for the isotropic. The error is

approximately six times greater for a steel beam than for beam of graphite-magnesium if there is a fixed

support. This result may be unique to these two materials, so other materials with both orthotropic and

anisotropic properties would need to be studied to determine if the error is always less for an anisotropic

solid. The results also show that this error difference does not exist for the simply-supported beam. For this

support case, the error values with or without the Poisson effect are virtually identical.

75
 5
Fixed−Free
Fixed−Fixed
4 Simply−Supported

Percent Error for Zero Poisson Ratio, %

3

−1
1 1.5 2 2.5 3 3.5 4 4.5 5
Mode of Vibration

Figure 4.17: Influence of Poisson Effect for Varying Modes of Vibration

4.3.4 Conclusions

From the results discussed, a number of conclusions can be made. As previously stated, these results are

only known for the geometry, materials, and constitutive laws that were used in the analysis. The determined

findings are listed below.

1. For beams with at least one fixed support the Timoshenko beam theory as well as an elasticity theory

with zero Poisson ratio underestimate the frequency results. For very slender beams the Euler-Bernoulli

and Rayleigh beam theories also underestimate resonant frequencies, which contradicts expectations.

2. For a simply-supported beam all three beam theories and the zero Poisson condition overestimate

frequency results. This occurs for all slenderness ratios and modes of vibration considered.

3. The slenderness ratio has a significant impact on the frequency accuracy for both the Euler-Bernoulli

and Rayleigh beam theories. As the slenderness decreases, the error increases rapidly. This problem

worsens for higher modes of vibration and anisotropic materials.

4. The resonant mode shapes for a non-slender beam deviate from one-dimensional beam theory predic-

tions with significant warping through the thickness and height.

76
 5. Neither slenderness nor anisotropy have a meaningful influence upon the frequency accuracy for the

Timoshenko beam model.

6. None of the three studied beam theories provide acceptable results when a fixed support exists. In

these cases, neglecting the Poisson effect leads to significant errors (more than 3%) for even extremely

slender beams. When analyzing beams with fixed support conditions, more complex analysis such as

the full elasticity approximation used in this study are necessary.

7. The influence of neglecting the Poisson effect decreases for higher modes of vibration with a cantilevered

or fixed-fixed beam.

8. For beams with only hinged or free supports the assumption of zero Poisson ratio has a minimal effect

on the frequency results. This is true for both orthotropic and isotropic materials as well as all of the

modes of vibration and slenderness ratios considered in this study.

The main purpose of this study was to determine when the typical beam theories are no longer acceptable.

To summarize the answer to this investigation, all three theories are inadequate for beams with at least one

fixed support regardless of slenderness or material type. For a simply-supported beam, the Timoshenko beam

model performs better than expected. The accuracy of frequency results only becomes poor for very low

slenderness values combined with high modes of vibration. The Euler-Bernoulli and Rayleigh beam theories

deteriorate in accuracy earlier than predicted. The common standard is that Euler-Bernoulli is acceptable

for slenderness ratios greater than 100, but the results of this study contradict this. Even with a slenderness

ratio of 138, the error for higher modes and orthotropic materials is near 2% for Euler-Bernoulli and above

1% for Rayleigh. The Timoshenko beam model is the clear choice for a simply-supported beam unless the

slenderness ratio is extremely high. For fixed-free or fixed-fixed beams, more complex analysis is required to

calculate accurate resonant frequencies and mode shapes.

4.4 Method for Investigating MEMS Sensor

The calculation work performed for a proposed new MEMS sensor determined accurate estimates of

resonant frequencies for certain modes of vibration. This device is planned to be an acoustic resonator with

possible applications in molecular detection, strain sensing in civil infrastructure, and electronic frequency

control. The primary sensing element of this device is a fixed-fixed bridge of silicon crystal with a thin film

of gold. An illustration of the device, cross-sectioned through the middle is provided in Figure 4.18. As the

77
sensor experiences strain or mass loading from certain molecules the resonant frequencies of this bridge will

shift and be detected. For the application is structural health monitoring the vision is that dynamic strains

from both large-scale events such as earthquakes and small-scale vibrations from acoustic emissions could

be detected and output. This device could also possibly be operated wirelessly, where the signal would be

directly transmitted to a processing node. This would greatly reduce the complexity and cost over other

sensor systems available.

Figure 4.18: Illustration of MEMS Sensor Configuration Cross-Sectioned Through the Mid-Length

To proceed further with the development of this device, the inventor, Dr. Ward Johnson, will determine

the exact resonant frequency range of the fixed-fixed bridge experimentally. Before this is feasible, more

accurate estimates of these frequencies are needed so that they can be found experimentally. For the purpose

of these estimates the thin gold film will be neglected and only the properties of silicon will be input. Due

to the results obtained from the vibration study discussed previously, the ritz-based approximation program

was utilized to calculate the resonant frequencies and mode shapes, since a higher error was found in all

three beam theories when a fixed support is involved.

A few different bridge dimensions were analyzed in this work, with the length of the bridge varying.

For all of the cases, the width of the bridge is 69 microns and the thickness was set to 4.9 microns. Three

different lengths were considered, since the best size in this dimension has not yet been decided. The three

lengths studied were 400 microns, 600 microns, and 800 microns. The elastic constants and density used

for the silicon material were determined by McSkimin and others and are provided in Table 4.8 [47]. These

material properties and dimensions were input into the already created Ritz-based approximation program

from the vibration study to determine resonant frequencies.

78
 Table 4.8: Material Properties for Silicon

Constant Value
C11 , GPa 167.4
C44 , GPa 65.23
C12 , GPa 79.57
Density, ρ, kg/m3 2331

The modes of vibration which are of particular interest for the operation of the device are the lateral-

flexural modes. The frequencies of lateral modes are more attractive because they experience less damping

which would lead to greater frequency resolution and sensitivity. The damping is lower for these modes than

for transverse modes because less air or liquid is displaced through the movement, and thus less energy losses

[48]. The Ritz-based approximation program was easily able to isolate the lateral-flexural modes of the bridge

by simply selecting proper axes orientation. The computing program uses set displacement patterns for each

axis which are based upon the assumption of vertical, or transverse displacement. By simply rotating the

dimensions of the silicon beam so that the width dimension was vertical, the lateral modes were found. The

determined resonant frequencies and mode shapes are discussed in the following section.

The developer of this device is also interested in determining the frequencies for the through-thickness

shear modes for the silicon bridge. These vibration modes would disturb less air than the lateral flexural

modes and would thus experience less damping and provide even greater frequency resolution. A simple

diagram of the cross section deformation for this type of vibration is displayed in Figure 4.19. The calculations

for the resonant frequencies of these mode types have not yet been completed. All other calculation work

and investigations for this thesis focus on flexural vibrations which have been studied in more depth, and

thus switching to shear mode calculations is a greater challenge than determining the lateral flexural modes.

The hope is that these calculations can be completed soon and provide another option for the development

of this new sensor.

4.5 Results from MEMS Sensor Investigation

Through the use of the Ritz-based approximation program for a fixed-fixed beam, the resonant frequencies

for the first five lateral-flexural modes were determined. The values for these results are provided in Table 4.9.

The fundamental lateral-flexural mode is of greatest interest for the development of the device but higher

modes were also calculated in case the need arose for these frequency estimations.

79
Figure 4.19: Illustration of Cross Section Deformation for Through-Thickness Shear Mode

Table 4.9: First Five Lateral-Flexural Mode Frequencies for MEMS Bridge
Length (microns) Mode Frequency (rad/s)
1 18225324
2 44972102
400 3 78724660
4 116570326
5 158118357
1 8548456
2 22215588
600 3 40562569
4 62399670
5 86980150
1 4908432
2 13058480
800 3 24389408
4 38414047
5 54548071

(a) 400 microns (b) 600 microns (c) 800 microns

Figure 4.20: Lateral-Flexural Vibration Modes for MEMS Bridge

80
 Three-dimensional mode shapes for the fundamental lateral-flexural modes were plotted for each of the

three lengths of interest. This was done to ensure that the frequencies calculated were for the correct vibra-

tional modes of interest. These mode shapes are shown in Figure 4.5. The calculated mode shapes followed

expectations and were confirmed to match the vibrational modes of interest for the MEMS sensor. With this

information, the hope is that exact resonant frequencies can more easily be determined experimentally from

the accurate estimations. This will further the development and testing process for the device which could

potentially provide significant applications and advantages in cell sensing and structural health monitoring.

81
Chapter 5

Conclusions

5.1 Summary of Work and Results

The bridges and buildings that the nation depends upon daily to provide safe and efficient usage, con-

tinually deteriorate from harsh environmental conditions and operational loads. Many of these structures

have been in service for decades and are at risk of significant damage and deficiency. The field of structural

health monitoring has arisen to attempt to prevent catastrophic failures and prioritize the immense amount

of repair and maintenance that is required through persistent monitoring of the strength and integrity of a

structure. Although many methods and technologies have been proposed and developed for the practice of

structural health monitoring, a need still exists for further improvements to increase the applicability and

affordability.

The research for this thesis aimed to contribute to possible advancements in the monitoring of civil

infrastructure through three main goals. The first of these was to provide a thorough review of the concepts,

methods, and technologies currently in use for structural health monitoring. This review presented the

objectives of this practice and explained the theory behind multiple methods which have been established

to fulfill these goals. Throughout these explanations, the benefits and drawbacks of each method were

discussed and compared. This survey also offered the details of numerous sensor options which are available

to obtain the data required to perform the analysis component of structural health monitoring. The concepts

utilized for each of these technologies, their application capabilities, advantages, and disadvantages were all

presented. The motivation behind this overall review was to provide a general understanding of the field

which could better direct further research. The primary motivation for this was to aid calculations included

in this thesis, but will hopefully guide others in their investigations and experiments as well.

82
 The second objective of this paper was to perform an in-depth study of vibrational beam mechanics

theories to determine when and if their accuracy is unacceptable due to three key assumptions employed in

all of them. This investigation was completed to ensure acceptable results could be obtained for calculations

of a possible new structural health monitoring sensor. Three of the most common beam theories were

considered: Euler-Bernoulli, Rayleigh, and Timoshenko. The Euler-Bernoulli model is the simplest option

since it ignores the influences of the rotational inertia and shear forces in a beams displacement. The Rayleigh

theory adds in rotary inertia, while Timoshenko includes both of these effects. All three of these theories

assume that the deformation occurs primarily in one dimension, that the beam is relatively slender, that

the material is isotropic, and that the Poisson effect is negligible. By employing an accurate Ritz-based

elasticity approximation, each of these assumptions was tested to determine how it affects the accuracy of

frequency and mode shape findings. Varying support conditions, beam lengths, and material properties were

used in the analysis to better isolate each of the effects of interest. The conclusions of this investigation were

compelling. A list of the key results found is provided below.

• When the slenderness ratio falls below 50 the errors in frequency results increases rapidly for both the

Euler-Bernoulli and Rayleigh beam models, with errors up to 90% in one case.

• For higher modes of vibration and a slenderness ratio of 138, the frequency errors are above 2% for

Euler-Bernoulli and Rayleigh, contradicting the assumption that these theories are adequate for beams

with s > 100. This loss of accuracy for higher modes was consistent for all three beam cases.

• When the Euler-Bernoulli or Rayleigh theories are applied to an orthotropic material the frequency

errors for s < 50 more than double in comparison to the errors for an isotropic material. This trend

was found for all three support cases.

• For all three support cases studied the frequency errors for the Timoshenko model are not affected by

very low slenderness or anisotropic properties.

• Due to neglecting the Poisson effect, none of the three beam theories are acceptable for beams with

a fixed support since the frequency errors are 4% or greater for the entire range of slenderness values

studied.

• When the influence of the Poisson ratio was isolated, it was found that neglecting this effect leads to

frequency errors greater than 3% for beams with s > 300 when a fixed support exists.

83
 • For the simply-supported case, the Poisson effect has a minimal influence and it was found that the

Timoshenko model gives excellent results for all slenderness ratios considered.

In summary, it was found that a more accurate, three-dimensional analysis should be utilized for beams

involving a fixed support, but when this is not present the Timoshenko model provides excellent results for

all slenderness values, materials, and vibrational modes investigated.

The final and motivating goal of this thesis work was to determine accurate estimates of resonant frequen-

cies for the sensing component of a proposed MEMS sensor. This sensing component consists of a fixed-fixed,

thin silicon bridge. Because of the support conditions of this bridge, the highly accurate three-dimensional

elasticity solution used in the vibrational study was employed for the calculations. With this option available,

accurate frequency estimates were obtained for the lateral-flexural modes of the bridge which would aid in

the experimental testing and development of this device.

5.2 Suggestions for Continued Research

Numerous opportunities exist for further research stemming from the work presented in this report. These

research possibilities develop from two of the main goals of this thesis. Firstly, there are multiple additional

aspects which could be investigated to expand upon the vibrational mechanics study in this paper. Another

assumption of typical beam theories is that the cross section of the beam is symmetrical, and it would be

interesting to investigate beam dimensions which do not follow this parameter. Varying beam dimensions,

in general, could be studied to better understand even the effects examined in this thesis. It is unclear if the

existence of a fixed support would have as significant of an influence for non-square cross sections, hollow

beams, and more which could be of interest for multiple applications. Another option for further study,

would be to include a greater variety of material properties to further understand the effect of anisotropy on

vibrational frequencies and mode shapes. More orthotropic materials could be included, as well as entirely

anisotropic materials. Since the inclusion of a fixed support clearly had a meaningful impact on the action of

a beam, a deeper investigation into the deformations, strains, and forces experienced at a fully fixed support

could provide meaningful information to better understand this anomaly.

Another direction that additional research could go from the vibrational study presented, would be to

employ the more accurate three-dimensional elasticity approximation to applications that involve modal

analysis and fixed supports. For example, there could be interest in how differently a structural frame

modeled with fixed connections responds to modal analysis when the Poisson effect is included. Also, many

84
MEMS devices which have been developed involve the response of a cantilevered beam, and the inclusion of

the Poisson effect in their analysis could affect the results. A multitude of similar scenarios exist where a

beam with a fixed support is involved which could benefit from applying the results found in this thesis.

Research possibilities could also follow the work done for the proposed MEMS sensor. Some research will

likely occur by the developers to experimentally determine the actual resonant frequencies of the device and

to decide which length for the silicon beam would provide the best operation. If the device has substantial

potential, further work would be required to determine its accuracy in actually monitoring civil infrastructure.

There is also potential for applicability in biomechanical fields such as cancer cell detection and more research,

testing, and development could also pursue this opportunity. In general, there are countless directions that

new investigations could follow from this report and the hope is that the work presented here will provide a

useful base to begin from.

85
Bibliography

[1] American Association of State Highway and Transportation Officials. Bridging the Gap: Restoring and

Rebuilding the Nation’s Bridges. 2008.

[2] U.S. Department of Transportation, Research and Innovative Technology Administration, Bureau of

Transportation Statistics. Transportation Statistics Annual Report 2010. Washington, 2011.

[3] Doebling, Scott W., Charles R. Farrar, and Michael B. Prime. “A Summary Review of Vibration-Based

Damage Identification Methods.” Shock and Vibration Digest. 30.2 (1998):91-105.

[4] Farrar, Charles R., Scott W. Doebling, and David A. Nix. “Vibration-Based Structural Damage Iden-

tification.” Philosophical Transactions of the Royal Society. 359. (2001):131-149.

[5] Doebling, Scott W., Charles R. Farrar, and Randall S. Goodman. “Effects of Measurement Statistics

on the Detection of Damage in the Alamosa Canyon Bridge.” 1997. 919-929.

[6] Chang, Peter C., Allison Flatau, and S.C. Liu. “Review Paper: Health Monitoring of Civil Infrastruc-

ture.” Structural Health Monitoring. 2.3 (2003): 257-267.

[7] Alvandi, A., and C. Cremona. “Assessment of Vibration-Based Damage Identification Techniques.”

Journal of Sound and Vibration. 292. (2006):179-202.

[8] Cruz, Paulo J.S., and Rolando Salgado. “Performance of Vibration-Based Damage Detection Methods

in Bridges.” Computer-Aided Civil and Infrastructure Engineering. 24 (2008):62-79.

[9] Chong, Ken P., Nicholas J. Carino, and Gleen Washer. “Health Monitoring of Civil Infrastructures.”

Smart Materials and Structures. 12. (2003):483-493.

[10] Gastineau, Andrew, Tyler Johnson, and Arturo Schultz. Minnesota Department of Transportation.

Bridge Health Monitoring and Inspection - A Survey of Methods. 2009.

86
[11] Aktan, A. Emin, F. Necati Catbas, Kirk A. Grimmelsman, and Mesut Pervizpour. Federal Highway

Administration Research and Development. Development of a Model Health Monitoring Guide for Major

Bridges. 2003.

[12] Phares, Brent M., Terry J. Wipf, Lowell F. Greimann, and Yoon-Si Lee. Wisconsin Department of Trans-

portation. Health Monitoring of Bridge Structures and Components Using Smart-Structure Technology

Volume 1. 2005.

[13] Park, H.S., H.M. Lee, H. Adeli, and I. Lee. “A New Approach for Health Monitoring of Structures:

Terrestrial Laser Scanning.” Computer-Aided Civil and Infrastructure Engineering. 22.1 (2007):19-30.

[14] Wilson, Jon, Craig Aszkler, and Timothy Geiger. “Acceleration, Shock, and Vibration.” Test and Mea-

surement: know it all. Boston: Elsevier Inc., 2009. 101-151.

[15] Introduction to Piezoelectric Accelerometers. 2012. Graphic. PCB Piezotronics, Inc. 2 Apr 2012.

<http://www.pcb.com/techsupport/tech accel.php>.

[16] Choi, Haksoo, Sukwon Choi, and Hojung Cha. “Structural Health Monitoring System Based on Strain

Gauge Enabled Wireless Sensor Nodes.” 5th International Conference on Networked Sensing Systems.

211-214. 17-19 June 2008.

[17] Zou, Y., L. Tong, and G.P. Steven. “Vibration-Based Model-Dependent Damage (Delamination) Identi-

fication and Health Monitoring for Composite Structures - A Review.” Journal of Sound and Vibration.

230.2 (2000):357-378.

[18] Kesavan, Ajay, Sabu John, and Israel Herszberg. “Strain-based Structural Health Monitoring of Complex

Composite Structures.” Structural Health Monitoring. 7.3 (2008):203-213.

[19] Laughton, Michael, Douglas Warne and E.A. Parr. “Strain Gauges, Load Cells, and Weighing.” Test

and Measurement: know it all. Boston: Elsevier Inc., 2009. 267-280.

[20] The Strain Gauge. 2007. Automatic Control, Robot and Mechatronics Labs of Mechanical Engineering

Department. Web. 3 Apr 2012. ¡http://web.deu.edu.tr/mechatronics/TUR/strain gauge.html¿.

[21] Watkins, S., J. Fonda and A. Nanni. “Assessment of an Instrumented Reinforced Concrete Bridge with

Fiber-Reinforced Polymer Strengthening.” Optical Engineering. 46.5 (2007).

87
[22] Ravisankar, K., S. Parivallal, T. Narayanan, K. Kesavan and R. Narayanan. “Vibrating Wire Strain

Gauges for Long-Term Monitoring of Structures.” Indian Concrete Journal. 75.8 (2001):535-541.

[23] Dally, J. and W. Riley. Experimental Stress Analysis. 4. Knoxville: College House Enterprises, 2005.

[24] Neild, S.A., M.S. Williams and P.D. Mcfadden. “Development of a Vibrating Wire Strain Gauge for

Measuring Small Strains in Concrete Beams.” Strain. 41.1 (2005):3-9.

[25] Domalik, D., J. Shura and D. Linzell. “Design and Field Monitoring of Horizontally Curved Steel Plate

Girder Bridge.” Transportation Research Record. (2005):83-91.

[26] Vibrating Wire Strain Gages. 2011. Photograph. CARBO Ceramics. 2 Apr 2012.

<http://www.carboceramics.com/agi/strain-gauges>.

[27] Corda, J. and L.K. Al-Tayie. “Enhanced Performance Variable-Reluctance Transducer for Linear-

position Sensing.” IEE Proceedings: Electrical Power Applications. 150. (2003):623-628.

[28] McDonald, P.C. and C. Iosifescu. “Use of LVDT Displacement Transducer in Measurements at Low

Temperatures.” Measurement Science and Technology. 9.4 (1998):563-569.

[29] Fischer-Cripps, Tony. “Position.” Test and Measurement: know it all. Boston: Elsevier Inc., 2009.

255-265.

[30] Tilt Measuring - Inclinometers - Levelling. 2012. Graphic. ALTHEN Mess - und Sensortechnik, Frank-

furter. 3 Apr 2012. <http://www.althensensors.com/inclinometer and tilt sensors/ait930>.

[31] Vallen, Hartmut. “AE Testing Fundamentals, Equipment, Applications.” NDT.net 7.9 (2002).

[32] Casas, J. and P. Cruz. “Fiber Optic Sensors for Bridge Monitoring.” Journal of Bridge Engineering. 2.3

(2003):257-267.

[33] Yin, S., P. Ruffin and F. Yu. Fiber Optic Sensors. 2nd edition. Boca Raton: CRC Press, 2008.

[34] Inaudi, D. and S. Vurpillot. “Monitoring of Concrete Bridges with Long-Gage Fiber Optic Sensors.”

Journal of Intelligent Material Systems and Structures. 10.4 (1999):280-292.

[35] Ozevin, D., D.W. Greve, I.J. Oppenheim, and S.P. Pessiki. “Resonant capacitive MEMS acoustic emis-

sion transducers.” Smart Materials and Structures. 15.6 (2006):1863-1871.

88
[36] Mohammed, A.A.S., W.A. Moussa and E. Lou. “A novel MEMS strain sensor for structural health

monitoring applications under harsh environmental conditions.” Structural Health Monitoring 2007:

Quantification, Validation, and Implementation. Ed. F.K. Chang. Lancaster: DEStech Publications,

Inc., 2007. 121-128.

[37] Rajagopalan, Jagannathan and M. Taher A. Saif. “MEMS sensors and microsystems for cell mechanobi-

ology.” Journal of Micromechanics and Microengineering. 21.5 (2011).

[38] Gao, Song and Xinyan Wang. “Microelectromechanical system-based diagnostic technology for cervical

cancer.” Journal of Cancer Research and Clinical Oncology. 137.11 (2011):1721-1727.

[39] Han, S.M., H. Benaroya, and T. Wei. “Dynamics of Transversely Vibrating Beams Using Four Engi-

neering Beam Theories.” Journal of Sound and Vibration. 225.5 (1999):935-988.

[40] Timoshenko, S.P. History of Strength of Materials. New York: McGraw-Hill Book Company, Inc., 1953.

[41] Timoshenko, S.P. “On the correction for shear of the differential equation for transverse vibrations of

bars of uniform cross-section.” Philosophical Magazine. 41. (1921): 744-746.

[42] Fung, Y.C., & Tong, P. Classical and Computational Solid Mechanics. Hackensack: World Scientific

Publishing Co. Pte. Ltd., 2010.

[43] Puchegger, S., D. Loidl, K. Kromp, and H. Peterlik. “Hutchinson’s Shear Coefficient for Anisotropic

Beams.” Journal of Sound and Vibration. 266. (2003):207-216.

[44] Heyliger, P.R., and A. Jilani. “The Free Vibrations of Inhomogeneous Elastic Cylinders and Spheres.”

Int. J. Solids Structures. 29.22 (1992):2689-2708.

[45] Ohno, Ichiro. “Free Vibration of a Rectangular Parallelepiped Crystal and its Application to Determina-

tion of Elastic Constants of Orthorhombic Crystals.” Journal of Physics of the Earth. 24. (1976)355-379.

[46] Ledbetter, H.M., S.K. Datta, and T. Kyono. “Elastic Constants of a Graphitemagnesium Composite.”

Journal of Applied Physics. 65:9. (1989):3411-3416.

[47] McSkimin, H.J., W.L. Bond, E. Buehler and G.K. Teal. “Measurement of the Elastic Constants of

Silicon Single Crystals and Their Thermal Coefficients.” Physical Review. 83.5 (1951):1080-1080.

89
[48] Heinrich, S.M., R. Maharjan, I. Dufour, F. Josse, L.A. Beardslee and O. Brand. “Analytical Model

of a Thermally Excited Microcantilever Vibrating Laterally in a Viscous Fluid.” IEEE Sensors 2010

Conference. 1399-1404. Nov 2010.

90