Você está na página 1de 142

Condition Monitoring Sensor for Reinforced

Elastomeric Materials
A Thesis submitted to the Graduate School
of the University of Cincinnati
in partial fulfillment of the
requirements of the degree of
Master of Science
in the Department of Mechanical Engineering
of the College of Engineering
by

Charles Matthew Dandino


B.S. in Mechanical Engineering
University of Cincinnati
June 2011

Committee Chair: Dr. Mark J Schulz


1

i.

Abstract

In-situ monitoring of materials is a great problem in the field of structural health monitoring. The ability
to receive real time data relaying the condition of a body is an elusive but invaluable goal. Even more
difficult is monitoring the continuous body rather than a small subset of discrete points which may or
may not represent the health of the whole body. The final challenge, specific to elastomeric materials,
is to develop a sensor capable of surviving a great deal of strain as the body bends, flexes, and stretches
during typical operation.
This thesis provides a solution to these problems by exploring the development and performance of a
continuous sensor skin. This skin has been carefully developed to survive the operational metrics of
steel reinforced hydraulic hoses. This thesis explores several avenues for development with a focus on
those which show promise in hydraulic hose applications. Several different theories of how the sensor
may operate are discussed in detail while the three most common failure modes are tested: puncture,
tear, and foreign object damage.

ii.

Acknowledgements

Dr. Mark Schulz For providing tireless guidance in research, coursework, and proposal writing. The
responsiveness, care, and genuine interest in my success have made this rigorous curriculum and
career path possible. Keeping Dr. Schulz as an advisor was the most significant factor in my choice to
stay at the University of Cincinnati and I have never regretted it.
Dr. Vesselin Shanov For support in lab activities and management. Also, for personal support over the
past 4 years as I progressed as a researcher.
Dr. Jay Lee For providing the industry connections and support staff necessary for this project to exist.
Ms. Alia Majid For providing a critical line of communication between the research team at UC and
the engineering division at Parker. Also for strict, concise management practices which helped
streamline the development process.
Dr. Xiangdong (Dawn) Zhu - For excellent guidance and management during the project. Many hour
were spent helping organize the project and working with those involved to improve our
organizational skills and presentation.
Mr. Pete Buca For believing in and investing in academia; specifically this project. Mr. Bucas presence
in meetings greatly increased the effectiveness and productivity through his incredible knowledge
and ability to link problems with available resources.
Ms. Stephanie Weiss For providing insight and samples from EPD hoses.
Mr. Adam Hehr For sticking through a very challenging year of coursework and being a consistent,
reliable resource to bounce highly technical ideas off of and for listening to periodic venting.
Mr. Patrick Brown For the remarkably quick procurement of necessary research items and
reimbursement checks.

Mr. Brad Ruff For technical support and remarkable lab management as well as regular comedic relief.
Mr. Doug Hurd For excellent advice driven by an inherent technical understanding and years of topnotch experience.
Parents - For many years of consistent support and challenges.

iii.
Symbol Word
Bending Radius
Burst

Permittivity

Pinhole

Strain

RC200

Strain Rate
Conductive Paint

SHM

Structural Health
Monitoring

Definitions and Symbols


Definition
Recommended minimum radius from the center point of the bend
arc to the inside of the hose.
Failure mode of hydraulic hose in which the material of the hose fails
suddenly resulting in a rapid loss of pressure as the hydraulic fluid
escapes.
The measure of the resistance that is encountered when forming an
electric field in a medium. In other words, permittivity is a measure of
how an electric field affects, and is affected by, a dielectric medium.
Failure mode of hydraulic hose in which a very small hole allows high
pressure hydraulic fluid to shoot out. Pinhole failures are small
enough that they are not always visible, but the pressure makes them
very dangerous.
Normalized measure of deformation in a body representing the
displacement between particles in a body with respect to a reference
length.
Rate of change of strain with respect to time.
A custom conductive paint developed by the Nanoworld lab for this
project. It consists of 1 part by volume Elmers No Wrinkle rubber
cement, 2 parts by volume of 635 Southwestern graphite particles,
and 2 parts by volume toluene.
The process of implementing a damage detection and characterization
strategy for engineering structures

Table of Contents
i.

Abstract ................................................................................................................................................. 2

ii.

Acknowledgements............................................................................................................................... 4

iii.

Definitions and Symbols........................................................................................................................ 5

1.

Introduction ........................................................................................................................................ 13

2.

1.1.

Problem Statement ..................................................................................................................... 13

1.2.

Condition Monitoring System (solution)..................................................................................... 13

1.3.

Thesis Organization ..................................................................................................................... 16

Structural Health Monitoring of Reinforced Elastomeric Structures.................................................. 18


2.1.

2.1.1.

Need for SHM of Reinforced Elastomeric Structures ......................................................... 18

2.1.2.

Current technologies / Patent Study .................................................................................. 21

2.1.3.

Previous Research from the University of Cincinnati ......................................................... 23

2.2.

3.

Background ................................................................................................................................. 18

Failure Modes ............................................................................................................................. 25

2.2.1.

Burst .................................................................................................................................... 25

2.2.2.

Pinhole ................................................................................................................................ 26

2.2.3.

Abrasion .............................................................................................................................. 27

2.2.4.

Kink...................................................................................................................................... 27

2.2.5.

Corrosion ............................................................................................................................. 28

Sensor Skin for Surface Structural Health Monitoring ........................................................................ 30


3.1.

Sensor Skin Theory ...................................................................................................................... 30

3.2.

Sensor Skin Models for Go/No-Go .............................................................................................. 30

3.3.

Sensor Skin Models for Predictive Modeling .............................................................................. 31

3.3.1.

Multi-layered Sensor ........................................................................................................... 31

3.3.2.

Piezoresistive Strain Sensor ................................................................................................ 33

3.3.3.

Piezocapacitive Strain Sensor ............................................................................................. 36

3.3.4.

Electrical Model for Sensor Skin ......................................................................................... 38

3.4.

Sensor Skin Development ........................................................................................................... 42

3.4.1.

Requirements ...................................................................................................................... 42

3.4.2.

Form Factor ......................................................................................................................... 44

3.4.3.

Electrode Design ................................................................................................................. 45


6

3.4.4.

Custom Dielectric Development ......................................................................................... 61

3.4.5.

Cost Analysis ....................................................................................................................... 90

3.5.

3.5.1.

Response to Pinhole Damage ............................................................................................. 97

3.5.2.

Response to Impact Testing ................................................................................................ 98

3.5.3.

Response to Bending Testing .............................................................................................. 99

3.5.4.

Response to Burst Testing................................................................................................... 99

3.5.5.

Response to Impulse Testing ............................................................................................ 101

3.5.6.

Response to Thermal Testing ............................................................................................ 103

3.6.

5.

6.

7.

Problems and Concerns ............................................................................................................ 105

3.6.1.

Intrinsically Safe Device .................................................................................................... 105

3.6.2.

Temperature Induced Embrittlement ............................................................................... 107

3.6.3.

False Trigger Due to Abrasion ........................................................................................... 108

3.6.4.

Strain to Failure ................................................................................................................. 109

3.7.
4.

Prototype Tests ........................................................................................................................... 96

Potential Solutions to False Positive ......................................................................................... 110

Home Made Devices ......................................................................................................................... 112


4.1.

Impact Tester ............................................................................................................................ 112

4.2.

Mandrel ..................................................................................................................................... 114

Electroactive Polymer Sensor ........................................................................................................... 116


5.1.

Electroactive Polymer Theory ................................................................................................... 116

5.2.

Electroactive Polymer Models .................................................................................................. 117

5.3.

Experimental Data .................................................................................................................... 119

Fiber Optic Sensor for Strain and Temperature Monitoring............................................................. 121


6.1.

Fiber Optic Theory .................................................................................................................... 121

6.2.

Implementation into a Hose ..................................................................................................... 122

6.3.

Problems and Concerns for Fiber Optic Sensors....................................................................... 123

Conclusions and Future Work ........................................................................................................... 124


7.1.

Functionality ............................................................................................................................. 126

7.1.1.

Proven ............................................................................................................................... 126

7.1.2.

Potential ............................................................................................................................ 126

7.2.

Thermoresistive Tests ............................................................................................................... 127

7.3.

Additional Piezoresistive Tests.................................................................................................. 127


7

7.4.

Life Testing ................................................................................................................................ 128

7.5.

Additional Applications ............................................................................................................. 128

7.6.

Energy Harvesting ..................................................................................................................... 129

8.

Works Cited ....................................................................................................................................... 130

9.

Appendix ........................................................................................................................................... 131


9.1.

Assembly Procedures (Chronological) ........................................... Error! Bookmark not defined.

9.2.

Test Procedures (Chronological) .................................................... Error! Bookmark not defined.

9.3.

Data Analysis Code .................................................................................................................... 131

9.4.

Data 133

9.5.

Data Acquisition System ........................................................................................................... 137

Table of Figures
Figure 1: Cross section and cutaway of hydraulic power hose. Image courtesy of Gates corporation, used
3/19/2012 http://www.gates.com/images/Gates/HydraulicHoseProfile2.jpg .......................................... 14
Figure 2: Floating mass transfer hose. Image courtesy of Parker Hannifin ............................................... 14
Figure 3: US 7,051,580 B1 May 30, 2006 ................................................................................................... 21
Figure 4:US 6,722,405 B2 April 20, 2004................................................................................................... 22
Figure 5:US 5,378,991 January 3, 1995 ..................................................................................................... 22
Figure 6: US 7,555,936 B2 July 7, 2009 ..................................................................................................... 22
Figure 7: Bending test bed. Image courtesy of Surya Sundaramurthy ...................................................... 24
Figure 8: Final iteration of the Clip-on Sensor designed by Surya Sundaramurthy. Image courtesy of
Surya Sundaramurthy. ................................................................................................................................ 24
Figure 9: Sensor skin prototype made by Surya Sundaramurthy. Image courtesy of Surya Sundaramurthy
.................................................................................................................................................................... 25
Figure 10: Hose burst failure. Picture compliments of Parker Hannifin accessed from
http://www.safehose.com/hfa_sub1.html on 3/20/2012 ......................................................................... 25
Figure 11: Hose Kinking. Image from 3/20/2012 courtesy of
http://hostedmedia.reimanpub.com/TFH/Step-By-Step/FH10SEP_BUYHOS_01.jpg ................................ 27
Figure 12: Example of abrasion damage on hose. Image from 3/20/2012 courtesy of
http://www.forconstructionpros.com/article/10523276/proper-hose-repair-cuts-downtime ................ 27
Figure 13: Corroded floating oil transfer hoses. Image borrowed from Alibaba.com on 3/19/2012
http://i00.i.aliimg.com/photo/v0/110573130/floating_oil_jetty_hose_pipes.jpg .................................... 29
Figure 14 Schematic of basic sensor skin design ........................................................................................ 30
Figure 15: Sensor Skin Failure Detection .................................................................................................... 31
Figure 16: Sensor Skin Multi-Layer Model .................................................................................................. 32
Figure 17: Schematic image explaining the response of a strain gauge to tension and compression.
Courtesy of lzantux of the WikiCommons .................................................................................................. 33
8

Figure 18: Grid Layout of resistance contacts for piezoresistive strain sensor .......................................... 35
Figure 19: Parallel Plate Capacitor model. Image courtesy of Wikimedia Commons. Author: inductiveload
.................................................................................................................................................................... 36
Figure 20: Increased pressure in hose resulting in thinning dielectric and increased surface area of
electrodes. .................................................................................................................................................. 37
Figure 21: Generalized Bridge Circuit. Courtesy of the Brown U. Physics Department wiki.brown.edu ... 37
: Piezor22esistive Sensor Skin Model.......................................................................................................... 39
Figure 23: Debonding Tape Illustration ...................................................................................................... 44
Figure 24: Super Shield and Rubber Spray Coating images courtesy of MG Chemical and Blue Magic
respectively ................................................................................................................................................. 45
Figure 26: Second Layer of Masking applied ................................................. Error! Bookmark not defined.
Figure 25: First Electrode Sprayed On ........................................................... Error! Bookmark not defined.
Figure 27: Completed Sensors partially separated ..................................................................................... 46
Figure 28: Masking Partially Removed........................................................... Error! Bookmark not defined.
Figure 29: Application of Pure Rubber Spray Dielectric ................................. Error! Bookmark not defined.
Figure 30: Surface resistivity vs thickness plot of Super Shield courtesy of MG Chemicals
http://www.mgchemicals.com/products/841.html accessed 3/21/2012 ................................................. 46
Figure 31:4 Probe test on Super Shield Sample .......................................................................................... 47
Figure 32: Schematic of 4 probe test courtesy of Vessels42 of the Wikimedia Commons ........................ 47
Figure 33: Four Probe data for Super Shield Spray on electrode ............................................................... 48
Figure 34: Cooling a Super Shield on paper backing sample with liquid nitrogen ..................................... 49
Figure 35: Bending a Super Shield on paper sample at -60C ...................................................................... 49
Figure 36: Resistance between electrodes before puncture ...................................................................... 50
Figure 37: Resistance between electrodes after puncture had occurred and puncturing item was
removed ...................................................................................................................................................... 50
Figure 38: Dynamic 4 probe test of a sample that performed well (did not crack) ................................... 52
Figure 39: Dynamic 4 probe test of a sample that that cracked. This was typical of the super shield spray
on electrode ................................................................................................................................................ 52
Figure 40: Super Shield sample strained to 10% showing severe cracking ................................................ 53
Figure 41: Photograph of experimental setup ............................................................................................ 54
Figure 42: Liquid Paint matrix with 250% graphite by volume strained at 10% ......................................... 55
Figure 43: Rubber Cement matrix with 250% graphite by volume strained at 10% .................................. 55
Figure 44: Main Effects Plot (Matrix Material: 1-Liquid Tape, 2-Rubber Cement ...................................... 57
Figure 45: Interaction Plot. A: Strain %, B: Matrix Material (1-Liquid Tape, 2-Rubber Cement), C:
Graphite Content % by volume ................................................................................................................... 58
Figure 46: Normal Plot of Residuals ............................................................................................................ 59
Figure 47: Contact angle of RC200 electrode paint on various rubber samples supplied by Parker
Hannifin ....................................................................................................................................................... 61
Figure 48: Ben Nye pure latex rubber unaffected after 30 minutes soaking in toluene. ........................... 62
Figure 49: Blue Magic pure rubber spray completely dissolved after 30 minutes soaking in toluene. ..... 62
Figure 50: Representative sample of RC200 and latex sensor before damage was induced. .................... 63

Figure 51: Representative sample of RC200 and latex sample after damage was induced by piercing the
sample with a razor blade and removing the razor blade. ......................................................................... 63
Figure 52: Schematic of out of plane tear failure. ...................................................................................... 64
Figure 53: Schematic of in plane tear failure .............................................................................................. 64
Figure 54: Sensor with an in plane tear failure. .......................................................................................... 64
Figure 55: Sensor with an out of plane tear failure. ................................................................................... 64
Figure 56: Unstrainted Latex Sample loaded in Instron Machine .............................................................. 65
Figure 57: Strained Latex Sample loaded in Instron Machine .................................................................... 65
Figure 58: RC200 Electrode Sample loaded, unstrained, in the Instron 5984 ............................................ 66
Figure 59: Stress Strain plot of RC200 Custom Electrode Materail ............................................................ 67
Figure 60: Stress Strain plot of Latex mixed with Body powder at varying concentrations ....................... 67
Figure 61: Stress Strain plot of Latex mixed with HX-CECO thickener at varying concentrations .............. 68
Figure 62: Paint brush with cured latex stuck to it as a result of premature, salt-induced curing............. 69
Figure 63: Tempera and Latex samples on foil ........................................................................................... 70
Figure 64: Tempera sample showing clean tear on foil tear test ............................................................... 71
Figure 65: Latex sample stretching across tear during foil tear test .......................................................... 71
Figure 66: Thickened latex samples. 1-3 Airbrushed, 4-6 Paintbrushed on to foil ..................................... 73
Figure 67: Brushed, thickened latex stretching over a tear in foil .............................................................. 73
Figure 69: Latex and tempera sample stretching across a tear in aluminum foil ....................................... 74
Figure 68: Latex and tempera samples airbrushed and paintbrushed onto foil ........................................ 74
Figure 70: Cut tear sample showing that the coating does not stretch across the tear ............................ 75
Figure 71: Sample loaded into vice and brought to a strained state .......................................................... 75
Figure 72: Representative sensor sample with RC200 electrode material and one coat of tempera paint
as a dielectric. ............................................................................................................................................. 76
Figure 73: Sensor skin samples with varying thicknesses of orange tempera dielectric paint. From left to
right there are 2, 3, 4, and 5 coats .............................................................................................................. 77
Figure 74: Microscope images of the different coating thicknesses .......................................................... 77
Figure 75: Several sensor skin samples with 5 coats of dielectric each. The samples are separated into
insulated and not insulated as some of the dielectrics were not sufficient to insulate the electrodes..... 79
Figure 76: Microscope image of 5 coat tempera sensor skin sample ........................................................ 79
Figure 77: 45x microscope image of sample 357 in the strained state. ..................................................... 80
Figure 78: 45x microscope image of sample 368, a 3:1:2 mixture of tempera, latex, and water
respectively ................................................................................................................................................. 81
Figure 79: Electrode paint wrinkling where it makes contact with 3:1:2 tempera : latex : water paint .... 81
Figure 80: Change in resistance between the two electrodes as a result of in-plane tear damage .......... 82
Figure 81: Resistance between the electrodes following the tear in plane damage ................................. 83
Figure 82: Tear in plane failure of sensor 383.1 ......................................................................................... 83
Figure 83: 3:1 tempera : water dielectric poured onto the sensor without being thinned. Progression of
pinhole damage (sensor 389.1) .................................................................................................................. 84
Figure 84: Parker Hannifin 3155R1 rubber samples in the oven for vulcanization .................................... 85
Figure 85: PH rubber dielectric after an out of plane tear failure .............................................................. 86
Figure 86: PH rubber dielectric after an in-plane tear failure..................................................................... 86
10

Figure 87: 3:1:2 Tempera Latex Water samples after vulcanization .......................................................... 88
Figure 88: 3:1:2 Tempera Latex Water sample before and after tear damage .......................................... 88
Figure 89: Microscope image of bubbles in dielectric on hose sample 411 ............................................... 90
Figure 90: Microscope image of bubbles in dielectric of hose sample 414 ................................................ 90
Figure 91: Microscope image of bubbles and tear in dielectric of hose sample 412 ................................. 90
Figure 92: Pinhole failure test progression on hose sample 411 ................................................................ 90
Figure 93: Schematic of electrostatic coating machine. Image courtesy of the National Engineering
Research Center for Nanotechnology, Shanghai, People's Republic of China accessed June 28, 2012 ..... 91
Figure 94: Resistance test showing that the PVC had insulated the surface of the sample before the
second layer of electrode was applied ....................................................................................................... 92
Figure 95: Sample being electrostatically coated with PVC ........................................................................ 92
Figure 96: Sample with steel probe inserted to connect the electrodes. Resistance increased. ............... 92
Figure 97: Resistance between electrodes of sample with PVC dielectric. Sample is poorly electrically
insulated...................................................................................................................................................... 92
Figure 98: Pressure test sequence used on the composite release paper samples ................................... 95
Figure 99: Microscope image of composite release paper against a black background ............................ 96
Figure 103: Schematic of Data Acquisition ................................................................................................. 97
Figure 100: Hose samples curing on mandrel ............................................................................................. 97
Figure 101: Applying electrode to hose samples ........................................................................................ 97
Figure 102: Applying dielectric to hose samples ........................................................................................ 97
Figure 104: Typical resistance vs. Time Plot for Pinhole Damage .............................................................. 97
Figure 105: Typical impact force data and resistance response of the sensor........................................... 98
Figure 106: Hose sample after impact ax test ............................................................................................ 99
Figure 107: Hose sample before impact ax test ......................................................................................... 99
Figure 108: Hose being bent while the resistance is measured ................................................................. 99
Figure 109: Parker Hannifin Burst Test Apparatus...................................................................................... 99
Figure 110: Burst hose following burst test .............................................................................................. 100
Figure 111: Data from burst test machine ................................................................................................ 100
Figure 112: Pressure and resistance values for burst test sample 20 ...................................................... 100
Figure 113: Pressure and resistance values for burst test sample 14 .................................................... 100
Figure 114: Impulse hose failure............................................................................................................... 101
Figure 115: Initial hose setup for impulse test ......................................................................................... 101
Figure 116: Long term data from hose impulse test showing machine shut-off and hose failure ........... 102
Figure 117: Short term data from hose impulse test showing pressure cycles and hose failure ............. 103
Figure 118: Sensor samples in the oven with wires attached for resistance measurements .................. 103
Figure 119: Temperature profile of the sensor samples .......................................................................... 104
Figure 120: Focus on induced failure data from thermal tests................................................................. 104
Figure 121: Full resistance data for thermal tests .................................................................................... 104
Figure 122: Zener diode circuit courtesy of Wikimedia Commons. Accessed 7/17/2012 ....................... 106
Figure 123: Determination of Tg by dilatometry. The linear sections below and above Tg are marked
green; Tg is the temperature at the point of intersection of the corresponding red regression lines.
Image courtesy of Afluegel ....................................................................................................................... 107
11

Figure 124: Strain induced on layers of materials with very different stiffnesses ................................... 109
Figure 125: Parker Hannifin's impact test apparatus ............................................................................... 112
Figure 126: University of Cincinnati's Impact Tester ................................................................................ 113
Figure 127: Mandrel constructed at UC.................................................................................................... 114
Figure 128: Electroactive polymer sensor courtesy of
http://lmts.epfl.ch/files/content/sites/lmts/files/shared/images/DEAP_principle.jpg Accessed 4/9/2012
.................................................................................................................................................................. 116
Figure 129: Superposition of capacitance in EAP sensors ........................................................................ 118
Figure 130: Fiber Bragg grating multiplexing. Image courtesy of Energy Harvesting for Structural Health
Monitoring Sensor Networks, Los Alamos National Lab, 2007................................................................. 122
Figure 131: Schematic of energy harvesting circuit courtesy of Los Alamos National Lab ...................... 129
Figure 132 NI 9219 Channel and Terminal Assignments .......................................................................... 138
Figure 133 NI 9219 Connections in 4-Wire Resistance and 4-Wire RTD Modes (reference: NI 9219
Operating Instructions and Specifications) ............................................................................................... 138
Figure 134 Data Acquisition System ......................................................................................................... 139
Figure 135 Measurement Circuit for Using NI 9219 ................................................................................. 140
Figure 136 Laptop IP Configuration .......................................................................................................... 141

12

1. Introduction
1.1.Problem Statement
Structural Health Monitoring (SHM) has proven valuable for monitoring the integrity of stiff structures
such as bridges but it fails to bring those benefits to critical, elastic components such as hydraulic hose,
tires, and drive belts. A tire blowout or hydraulic hose failure may result in reduced productivity,
product and equipment damage, or loss of life. Traditional SHM techniques use data from
accelerometers, strain gauges, fiber optic systems, and active piezoelectric patches to detect damage in
stiff structures and rigid components. (1) However, the unique coupling of a highly flexible, deformable
material, cord reinforcement, and a pressurized fluid (oil, air, etc.) confuses the traditional SHM
techniques. It is difficult to distinguish damage signatures from normal sensor responses. Finite element
analysis is currently difficult and unreliable for modeling reinforced elastomers (2). Furthermore, stress
concentrations arise as a result of significant stiffness differences between the sensor and structure
which jeopardize the structural integrity and sensor effectiveness. New sensors and techniques must be
developed to extract useful information regarding the unique failure modes of elastic components such
as pressure damage, pinhole puncture, rubber deterioration, and reinforcement cord failure.

1.2.Condition Monitoring System (solution)


The condition monitoring system proposed in this thesis focuses on rubber hydraulic hoses reinforced
with steel banding. The two sub categories of hoses are 1) hydraulic power hoses used to transfer
mechanical energy and 2) fluid mass transfer hoses used for transferring material. The primary
difference between these hoses is that the power transfer hose operates are a much higher pressure
and employed braided steel cable as a reinforcing agent as shown in Figure 1. The fluid transfer hose
operates at a much lower pressure and uses a helical steel cable for reinforcement as shown in Figure 2.

13

Figure 1: Cross section and cutaway of hydraulic power hose. Image courtesy of Gates corporation, used 3/19/2012
http://www.gates.com/images/Gates/HydraulicHoseProfile2.jpg

Figure 2: Floating mass transfer hose. Image courtesy of Parker Hannifin

A common element of these hoses is the potential for leak. A leak can occur anywhere along the hose
and visual detection methods may be inadequate in many applications. For example, many power
transfer hoses are behind a cover to reduce abrasion and environmental wear. This makes leaks
undetectable by visual inspection. Mass transfer hoses are often used to transfer goods from an
14

offshore rig to a ship or from ship to ship. These hoses are partially or entirely submerged in water
which necessitates divers for visual hose inspection. Such methods are inefficient and may not detect a
leak before significant damage is done to the product or environment.
To monitor the entire length of the hose for leaks, a flexible sensor skin has been proposed. This
concept was first proposed by Surya Sundaramurthy in 2011 (3) . The sensor skin concept has several
advantages:

Simplicity

Low power consumption

Monitors entire hose (rather than localized sensing)

Another condition monitoring methodology is to monitor the use and condition of the hose. That is,
monitor the stress/strain cycles from bending or pressure cycles and the environment in which these
conditions occur. Stress strain cycles are commonly used to determine remaining life of a structure, but
these techniques are often limited to metals which are relatively unaffected by strain rate and
temperature. Polymers, on the other hand, are highly sensitive to strain rate and temperature as shown
by Andrews and Tobolsky in 1952 (4). The time/temperature relationship can be quantified by the
equation below known as Williams-Landel-Ferry equation:
(

)
(

To take these affects into account, both temperature and strain must be measured with respect to time
at many locations along the hose. Several options for these measurements will be discussed later.

15

Advantages

Promise for accurate deterioration monitoring

Promise for deterioration modeling

Ability to perform preventative maintenance based on predictive models.

Disadvantages

Complicated

Expensive

Relatively high power consumption

1.3.Thesis Organization
This thesis begins at Chapter 2 with an explanation of structural health monitoring (SHM) as it relates to
reinforced elastomeric structures. The need for SHM is explored as well as the current state of the art
available to address some of these problems. The thesis then focuses on hydraulic hoses by focusing on
the various failure modes to be detected.
Chapter 3 explores a sensor skin concept as a potential solution. The theory of the sensor skin as well as
several models for how to use the sensor are presented. The chapter then continues to explore the
development of a functional sensor skin prototype. This development includes developing custom
electrode and dielectric paints. Finally, the problems and concerns incorporated with this type of sensor
are explored in detail.
Chapter 4 explains the custom tools used to apply the sensor skin to the samples and test the samples
for foreign object damage. These custom developed tools have proven highly valuable through the
course of the work.

16

Chapter 5 explores an electroactive polymer sensor. The chapter explains the theory of the
electroactive polymer (EAP) and explores models of how to use the EAP sensors in a hose for practical
purposes.
Chapter 6 explores fiber optic sensors. The chapter explains the theory of the fiber optic sensor and
explores methods of implementing such a sensor into a hydraulic hose. The chapter concludes by
addressing problems and concerns with the fiber optic sensors.

17

2. Structural Health Monitoring of Reinforced Elastomeric Structures


2.1.Background
2.1.1. Need for SHM of Reinforced Elastomeric Structures
The failure of a hydraulic hose can be disastrous. Power transfer hoses carry hot oil at pressures up to
6000 PSI for Parker Hannifins 792ST-12 hose (5). If such a hose fails it can result in a jet of oil at a
pressure high enough to inject corrosive hydraulic oil into the bodies of workers who come in contact
with the jet. Alternatively, the oil may catch fire and cause damage to personnel or equipment. Several
of these incidents have been recorded by the Occupational Safety and Health Administration (OSHA).

18

OSHA Accident Reports (6)

10/19/2000
Accident: 171059991 - Employee Burned When Hydraulic Fluid Ignites

Employee #1, an inspector at a pipe casting plant, was operating a pressure testing station on pipes of
various sizes. A hydraulic hose ruptured next to the operators station, atomizing fluid at 1,200 psi and
spraying it directly at a gas-fired floor heater next to Employee #1. The cloud of hydraulic fluid ignited,
and a large fireball engulfed an area of approximately 20 square feet. Employee #1 sustained severe
burns on his back, ears, arms, and hands.

9/2/1998
Accident: 170764609 - Employee Injured When Struck By Pressurized Hydraulic Fluid

Employee #1 was repairing and testing a hydraulic pump in the maintenance area of a sawmill. He had
his hand over the open valve to the pump on the outflow side when it was struck by a high pressure
stream of hydraulic fluid. The stream punctured his hand and forced hydraulic fluid inside, resulting in
injuries that required hospitalization. The causal factor in this accident was an open hydraulic line
capable of developing a pressure of 2,000 psi.

12/04/1996
Accident: 200270148 - Employee Killed When Struck By Falling Lift

Employee #1 was running a new wire to a material lift table. The table had been raised using a forklift
and was blocked to prevent it from falling. Once he had finished wiring the hydraulic motor, he removed
19

the supporting blocks. The lift fell, striking Employee #1 on the head. He was killed. The lift dropped
because there was no hydraulic fluid in the back cylinder.

2/19/1996
Accident: 170587760 - Two Employees Burned In Flash Fire

Employees #1 and #2, both press operators, were operating an aluminum extrusion press when a
hydraulic hose developed a leak. Hydraulic fluid sprayed out in a small stream under pressure and then
partially vaporized. The heat and flame from a nearby oven ignited the fluid, resulting in a flash fire.
Employees #1 and #2 sustained second- and third-degree burns.

6/14/1984
Accident: 14539035 - Bulldozer Operator Burned; Sprayed With Hot Hydraulic Oil

At approximately 2:45PM Employee #1 was getting ready to push over a tree using a Caterpillar D-8-H
bulldozer. As her operated the lever to raise the blade, a hydraulic hose located near the top left of the
blade which provided the force to move the blade burst. The ruptured hose sprayed hydraulic fluid back
along the left side of the diesel engine and into the operators location. Employee #1, the operator, was
struck by the hot hydraulic fluid, some of which had been ignited by contact with the exhaust manifold
and turbo charger. Employee #1 saw the hose burst as was able to close his eyes before being struck.
He stood up and jumped from the left running board to the ground, sustaining minor cuts when he fell
onto the rocky ground. He received first, second and third degree burns. He has undergone surgery
(skin grafting) and will require still more operations. He remains in serious condition with a 50/50
chance of survival.

20

2.1.2. Current technologies / Patent Study


This section will provide an overview of what technology already exists that relates, in some way, to this
research effort. Full patents available in the appendix.

US 7,051,580
The first design, below, is a double walled
apparatus that is partially filled with fluid. A
decrease in the fluid, as indicated by a fluid
sensor, indicates an external wall failure. An
increase in fluid indicates a failure in the
hydraulic hose.

Figure 3: US 7,051,580 B1 May 30, 2006

US 7,987,874
The next system is a double walled hose system where a pressure
sensor is located in the annulus to indicate increase in pressure
and, thus, a leak in the inner hose. This is for use with fluid having
a pressure significantly higher than atmospheric pressure as
stated in the patent text.

21

US 6,722,405
This system has high promise in the area
of damage control, but does not provide
damage detection before catastrophic
failure. A mechanical hydraulic hose valve
closes in response to a change in the hose
Figure 4:US 6,722,405 B2 April 20, 2004

length. The internal cable will pull the


valve closed, halting additional loss of material.
US 5,378,991
This design utilizes a multi-layer wall
construction on a rigid fluid vessel.
Alternating conducting and nonconducting material providing the means
for electrical connections based on
Figure 5:US 5,378,991 January 3, 1995

intrusion of fluid into the inner wall.

US 7,555,936
This system involves predicting structural failure
of a wall of a fluid containment vessel using a
strain-sensing means between the innermost and
outermost layers. This requires at least one
conductor parallel to the innermost layer of the
wall. The system utilizes changes in an electrical
Figure 6: US 7,555,936 B2 July 7, 2009

22

property resulting from distortion of the wall of the vessel causing distortion of at least one conductor
and therefore a change in capacitance between the conductive layers.
2.1.3. Previous Research from the University of Cincinnati
This problem is not a new problem. It has been investigated by talented researchers at the University of
Cincinnati for a few years prior to this thesis work. These investigators looked closely at five different
types of sensors as described below in the sensor road map prepared by Surya Sundaramurthy.
Function
Clip-on Sensor

Sensor for In situ


Monitoring and
Manufacturing
Quality Control

Sensor Type

Advantages

PVDF Piezo Film

Flexible, Long Life,


Cost $10, No
bonding required

Strain Gage

Local Monitoring
at hot spots

Wave Propagation

Simple
measurements

Impedance
Measurement

Simple
measurements

Eddy current
(Cylindrical Probe)

Non-contact

Disadvantages
Low operating
temperature,
Leads make it hard
to build-in.
Short lifetime,
Bonding
Hose needs to be
electrically
insulated
Hose needs to be
electrically
insulated
Time consuming

Results
On-line counter
built using
LabVIEW.
Debonded due to
surface
unevenness
Can detect surface
damage
Can detect
discontinuities and
pinhole damages
Detects surface
damage

The first two sensor types, PVDF clip-on sensor and strain gage were simultaneously tested on a bending
test bed. That testing apparatus is shown below.

23

Figure 7: Bending test bed. Image courtesy of Surya Sundaramurthy

After much experimentation, Surya Sundaramurthy's sixth design iteration was a rugged Clip-on Sensor
which incorporated a long, flexible PVDF sensor, a Mylar tape material, and a belleville washer (which
enables constant pressure load on the sensing segment).

Figure 8: Final iteration of the Clip-on Sensor designed by Surya Sundaramurthy. Image courtesy of Surya Sundaramurthy.

The next sensor type that was explored by Surya Sundaramurthy was damage detection through wave
propagation. A sensor/actuator was attached to each end of the hose. One end propagates a wave
24

through a hose while the other end reads the resultant vibrations. Wave responses of damaged and
undamaged hoses are recorded and used as benchmarks for determining the structural health of the
hose.
The final sensor design explored was a sensor skin. The sensor skin was comprised of two electrode
separated by a dielectric. The electrodes close a circuit when it is damaged because the dielectric
breaks down and the electrodes make contact.

Figure 9: Sensor skin prototype made by Surya Sundaramurthy. Image courtesy of Surya Sundaramurthy

2.2.Failure Modes
2.2.1. Burst
Hydraulic hoses may fail by bursting. This is the
most obvious failure mode as it is easily visible and
will result in a very sudden loss of pressure in the
hose. This can be differentiated from an abrasion
failure by examining the outer rubber and

Figure 10: Hose burst failure. Picture compliments of Parker


Hannifin accessed from http://www.safehose.com/hfa_sub1.html on 3/20/2012

reinforcement cords. If no sign of wire on wire or

25

cover abrasion is evident and the wires are broken randomly it is likely that the failure is a burst failure.
(7)
Burst failures indicate that the impulse cycle life of the hose has been exceeded. (7) Each pressure
impulse on the hose induces a stress and strain cycle on the reinforcing cables. These cycles fatigue the
material and eventually lead to failure. In the presence of pressure cycles, failure is most likely to occur
during a high pressure cycle and thus result in a burst (or sudden release of pressure). A burst hydraulic
hose may spray hot, flammable hydraulic fluid onto workers (resulting in burn injury) or hot equipment
(resulting in fire).
2.2.2. Pinhole
A pinhole failure is a unique and especially dangerous failure mode for hydraulic power transfer hoses.
Pinhole leaks can result from age, incompatible fluids, hose twist, and minimum bend radius violations.
(8) As the name suggests, pinhole leaks are very small and may not be visible. A leak can be indirectly
detected by the presence of hydraulic fluid where it should not be (ie. wet hose or small puddle).
If a pinhole leak is suspected, it is important to keep hands and limbs away from the affected area.
While the leak may appear harmless, the high pressure can inject caustic hydraulic fluid into the body of
the worker in similar fashion to a hypodermic needle. This injury may appear minor at first, but
hydraulic fluid can quickly destroy tissue, leading to gangrene and the need to amputate.
If a pinhole leak is suspected, the operator should check for the leak using a piece of cardboard as a
probe. Move the cardboard around the suspected area. If the cardboard become wet the leak has been
found.

26

2.2.3. Abrasion
Abrasion is the most common type of damage. Abrasion damage is the result of hoses rubbing against
each other or against their surroundings. (9) Abrasion damage is an external failure and should be
evident by visual inspection. Some abrasion is acceptable, but deep marring is cause for concern and if
any reinforcement braids are visible the hose should be replaced. Exposed reinforcement braids will
allow moisture to contact the steel braids and rust will begin to form. Weakened hose can lead to burst
failures.
If a large amount of abrasion is
visible, the hose should be replaced
immediately. If slight abrasion is
present, care should be taken to
remove the source of the abrasion
(ie. Separate rubbing hoses or move

Figure 12: Example of abrasion damage on hose. Image from 3/20/2012


courtesy of http://www.forconstructionpros.com/article/10523276/proper-hose-repair-cuts-downtime

hoses away from other components


which may cause rubbing).
2.2.4. Kink
A kink in a hose is the result of either a localized
force which will collapse the cylindrical walls of
the hose or an attempt to bend the hose less
than its minimum bending radius. In a typical
hose, a kink can be spotted by the loss of
cylindrical shape of the hose. At the point of the
kink, the hose shape will flatten and the two

Figure 11: Hose Kinking. Image from 3/20/2012 courtesy of


http://hostedmedia.reimanpub.com/TFH/Step-By-Step/FH10SEP_BUYHOS_01.jpg

27

opposing sides of the hose all will be parallel. The corners of the kink will have a very tight radius.
This tight radius at the corners produces a very high stress concentration which will fatigue the rubber
and reinforcing cable. Kinks can contribute to cracked rubber or separated reinforcement cords which
may result in burst or pinhole leak.
Kinks should be avoided by allowing enough room for the hose to bend without exceeding the bending
radius and by providing additional support for hoses that are bent around corners which are tighter than
the bending radius. Hoses should also never have a load applied radially inward as this may also result
in a kink.
2.2.5. Corrosion
Corrosion is the result of chemical changes in the material of a hose. These changes could come from
reactions with environmental elements such as salt or water to produce oxidation (rust) or from
reactions with ultraviolet (UV) rays from sunlight which will deteriorate many rubbers. Corrosion can be
detected by a discoloration of the hose, the presence of many small cracks (dry rot) or by a change in
the flexibility of the hose. Corroded hoses will be stiffer and more brittle than fresh hoses.
Corroded rubber will have an increased tendency to crack, flack off, and allow leaks. Corroded metal is
brittle, weak and much more likely to fail than metal that is not corroded.

28

Figure 13: Corroded floating oil transfer hoses. Image borrowed from Alibaba.com on 3/19/2012
http://i00.i.aliimg.com/photo/v0/110573130/floating_oil_jetty_hose_pipes.jpg

29

3. Sensor Skin for Surface Structural Health Monitoring


3.1.Sensor Skin Theory
The theory behind the sensor skin is elegantly simple. The theoretical functionality of the sensor skin
can be broken down into two parts: go/no-go indicator and a fatigue sensor as discussed below.
However, the primary benefit of this type of sensor is the skin design. This design enables the entire
surface of the hose to be monitored. In many cases system damage can be determined by a single local
sensor or sensor system (10). However, elastomeric structures (especially reinforced elastomeric
structures) have complex resonance patterns and cause severe vibration attenuation through the body.
These structures do not follow the same dynamics as most rigid body systems.
The ability to monitor the entirety of the structure simultaneously is invaluable in highly critical
applications.

3.2.Sensor Skin Models for Go/No-Go


The go/no-go model makes use of two flexible electrodes separated by a dielectric as shown in Figure
14.
Electrode
Dielectric
Electrode
Monitored Surface

Figure 14 Schematic of basic sensor skin design

In the event of a burst, pinhole, or abrasion failure the electrodes till push through the dielectric and
complete a circuit between them. The resistance between the two electrodes will go from effectively
to a relatively small, measureable value. Damage to the sensor will deform the electrodes as
demonstrated in Figure 15.

30

Impact

Electrodes

Dielectric
Pressure/fluid
Figure 15: Sensor Skin Failure Detection

3.3.Sensor Skin Models for Predictive Modeling


3.3.1. Multi-layered Sensor
The multi-layered sensor uses a simple go/no-go skin sensor at multiple locations along the thickness of
the hose wall. Feedback from the sensors will be relayed to a control unit independently to determine
where the damage has occurred. As a simple example, there may be a sensor on the inside of the hose
and on the outside. If both sensors fail simultaneously it most likely indicates a catastrophic failure that
has entirely ruptured the wall of the hose. However, if damage is only detected on the exterior of the
hose then the integrity of the hose is still intact. This exterior damage may be the result of abrasion or
sudden impact damage. The hose should be inspected soon, but can still be used. Damage detected
only on the interior of the hose may indicate that a pinhole leak is beginning to form, but has not broken
through the hose.
The location of these sensors can be as simple as interior/exterior as described above or could be
strategically located to inform the operator exactly what areas of the structure have been damaged.
This information can give a skilled operator the knowledge necessary to plan maintenance events and to
shut down the equipment before catastrophic failure occurs.

31

The example below shows a hose setup with sensors at three locations. A tear or puncture at each
sensor could indicate failures as listed in the table that follows.

Environment / Atmosphere
C
Sensor Skin
Layers

Reinforcement
Cable

A
Internal Fluid
Figure 16: Sensor Skin Multi-Layer Model

Sensor Reading
A
Open
Short

B
Open
Open

Open

Short

Open

Open

Open

Short

Short

Short

Open

Short

Open

Short

Open

Short

Short

Short

Short

Short

Possible Scenario

C
Open
Open

Hose is in good health. No significant damage.


Some damage to the interior of the hose. Fluid may be wearing at the
rubber.
Only damage detected within the hose, near the reinforcement.
Reinforcement may be fraying or broken to puncture sensor.
Exterior sensor damaged. Likely form sudden impact or abrasion on
the exterior of the structure.
Two internal sensors are damaged. This may indicate that a pinhole
leak is beginning to form. The structural integrity may be jeopardized.
The internal and external sensors are damaged. The most protected
sensor, B, is still undamaged. This may indicate general wear as the
inner most sensor closest to the reinforcement is still protected.
The two sensors surrounding the reinforcing cable are damaged. This
may suggest a problem with the reinforcement cable. Perhaps the
cable has been ruptured.
All sensors are damaged. The hoses has formed a leak or failed
catastrophically.

32

3.3.2. Piezoresistive Strain Sensor


Every sensor skin model and design involves some type of
sheet semiconductor. The piezoresistive effect is the
change in resistance in response to a change in the shape
(or strain) of a semiconductor. This effect is widely used in
standard strain gauges. The piezoresistive effect is
simplest to explain in the context of a strain gauge because
the piezoresistive effect in a strain gauge is due to axial
strain based on the following equation:

Figure 17: Schematic image explaining the response


of a strain gauge to tension and compression.
Courtesy of lzantux of the WikiCommons

Where

From a conductive materials perspective, the change in resistance in the presence of strain in the axial
direction of a wire can demonstrated by the conductivity equation of a wire.

Where

33

Excluding rare, auxetic materials, the cross sectional area of a material will decrease as the material is
strained in the direction normal to the cross section plane. The relationship between the deformation in
the strain direction and orthogonal to the strain direction is known as Poissons ratio. While the actual
value of Poissons ratio is dependent on the material, it is positive for most materials. Therefore, as the
length ( ) increases, the cross sectional area ( ) decreases. Since these properties are in the numerator
and denominator, respectively, the resistivity of the material will increase as it is strained axially.

34

A single conductive sheet plane could have a grid of contact points as shown in the figure below.
Multiplexing the resistance measurements between the conduction points could give a complete strain
profile. For example, in the figure below, measuring between the #A and #B points would show the
strain change between the top left and top right conditions. Measuring between #A and #D would
indicate a strain change from the top left and bottom middle conditions. Changes at different intervals
could produce a full strain map of the material surface.

1A

1B
2A

1A

1B

3A

3B

3D

3C

2D

2B

2A

2B

3A

3B

3D
2D

2C

1D

1C

1D

1A
2B
3A

3B

3D

3C

2D

2C

1D

1C

Figure 18: Grid Layout of resistance contacts for piezoresistive strain sensor

35

2C
1C

1B
2A

3C

This mapping quality, of course, is dependent on a predictable (preferably linear) relationship between
strain and resistance. A significant (same order of magnitude) temperature/resistance correlation may
drown out useful strain data. Similarly, unpredictable resistance response would render this method
useless.
3.3.3. Piezocapacitive Strain Sensor
The piezocapacitive strain sensor model uses the two
conductive sheets of the sensor skin as the plates of a parallel
plate capacitor as shown in Figure 19. The capacitance of this
type of capacitor can be calculated by the simple equation:

Where is permittivity,

is the area of each of the plates, and

Figure 19: Parallel Plate Capacitor model.


Image courtesy of Wikimedia Commons.
Author: inductiveload

is the distance separating the plates (in this case, the thickness of the dielectric). The parallel plate
model is valid for this case because the thickness of the sensors is 0.010 to 0.030 (including both
electrodes and the dielectric) and are made for hose applications with diameter ranging from 1 to 36.
Because the diameter of hose is much greater ( ) than the distance between the plates, this simplified
model will be a sufficient approximation.
Similar to the piezoresistive model in section 3.3.2, this model relies on strain to change the form of the
device and induce a detectable change in capacitance. In a hose, the area, A, of the capacitor would
increase when the hose is under pressure due to wall expansion. Simultaneously, the thickness of the
dielectric would decrease as its volume is stretched to a larger area. An increase in area and decrease in
thickness would cause the capacitance to increase by the equation above. The following figure
illustrates this simple change.

36

Pressure
Increase

Protective
Cover

Dielectric
Electrodes
Figure 20: Increased pressure in hose resulting in thinning dielectric and increased surface area of electrodes.

In many cases, the change in capacitance may be very small. There are two options for detecting a small
change in capacitance: 1) use a very sensitive measurement device or 2) amplify the signal through a
circuit such as a wheatstone bridge. Since sensitive measurement devices tend to be expensive and
protected by IP, it is much more interesting to discuss the wheatstone bridge. A generalized bridge
circuit is shown below.
If the signal source is allowed to be an alternating
current (AC) source, any component can be used to
create impedances Z1 through Z3. For our example, it
is convenient to assign Unknown Z to the impedance
measured across the two electrodes of the sensor (as
a capacitor) and to assign Z1 to a known capacitance.
Z3 can be a resistor of known resistance and Z2 can

Figure 21: Generalized Bridge Circuit. Courtesy of the Brown


U. Physics Department wiki.brown.edu

be a variable potentiometer to tune the bridge for balance measured at D. Below are the impedance
relationships used in an alternating current circuit.

37

Where is the imaginary number

is the frequency,

is the capacitance and

is the resistance.

This model is relatively simple for the case of strain induced exclusively from radial pressure; however, it
becomes more complicated when bending or twisting of a hose is taken into account. The current
model of the sensor skin is that of a skin which continuously covers the surface of the hose. A
continuous sensor will increase capacitance in some parts and decrease capacitance in others as the
hose strains randomly. It would be impossible to determine localized strain with a continuous sensor.
Additionally, if the primary function of the sensor skin (as a go/no-go sensor for puncture or tear
detection) was activated and the two electrodes were shorted the capacitance would drop to zero as
two parallel plates not isolated by a dielectric do not form a capacitor.
3.3.4. Electrical Model for Sensor Skin
The sensor skin can be modeled as an electrical circuit. The simplest model is to view it was a switch
with two resistors on each side. The resistors represent the flexible electrodes and the switch
represents the dielectric. When the dielectric is intact, the switch is open. Once the dielectric is
damaged, the switch is closed. This is the go/no-go model.

Electrode

Dielectric

Electrode

While this is the simplest model, tests have suggested that this model is inaccurate. In practice, the
dielectrics used in the paint on sensor skin behave like a relatively high impedance resistor. A more

38

accurate model of the sensor skin is three resistors in series. When the high impedance dielectric is
damaged, the lower impedance electrodes make contact and bypass the dielectric reducing the
impedance of the sensor circuit. Furthermore, bending tests have indicated that the circuit behaves as a
piezoresistive element. The primary source of this resistance change is assumed to be in the dielectric in
the following model. This piezoresistive effect may have application for pressure cycle counting and
strain detection.
Electrode

Dielectric

Electrode
:
Piezor22esisti
ve Sensor
Skin Model

However, further tests with AC current have suggested an even more complex model. . The inductance
and capacitance of the sensor was measured across the two electrodes at two different frequencies
using a BK Precision 878A LCR Meter. The samples showed a noticeable inductance and capacitance.
The inductance and capacitance values that the meter read changed based on the frequency of the
measurement (120 Hz versus 1 kHz). Since the meter reads both capacitance and inductance for the
sensor and the values change based on the frequency of the measurement it can be concluded that the
sensor, when applied to the hose, has inductance, capacitance, and resistance characteristics coupled
together.

39

SN
516
517
518
519
520
521
522
523
524
525
526
Resistor
(21k)

L @ 1kHz
(H)
14.9
15.3
0.0023
15
17
15.7
19.1
13
12.6
19.2
17.1

L @ 120Hz
(H)
445
631
0.0015
403
586
477.6
645.2
366.7
355.2
604
665

C @ 1kHz
(nF)
1.24
1.51
0.64
1.2
1.26
1.26
1.13
1.28
1.47
1.09
1.32

C @ 120Hz
(nF)
2.54
2.24
1.2
2.5
2.17
2.4
1.94
2.9
3.05
1.95
2.05

0.045

0.265

0.014

0.03

Electrode

Dielectric

Electrode

Further work needs to be done to determine what affects each aspect of this electrical model.
Resistivity
The resistivity of the sensor is likely a combination of three factors: electrode composition (polymer
matrix, conductive filler material, and conductive filler concentration), dielectric composition, and
physical dimensions. It may be possible to model the dielectric material as a homogeneous material
that follows the following formula.

40

Where R is the resistance,

is the resistivity, L is the length, and A is the cross sectional area. However,

when modeling the electrode material, it is more likely that percolation theory will need to be applied.
Changing the electrode composition, dielectric composition, and physical dimensions and running
additional tests would be necessary to determine the effects of each.
Capacitance
The sensor, when applied to a hose, can be modeled as a parallel cylinder capacitor. Adjusting the
physical dimensions (diameter, thickness, and length) of the sample should create trends in the data. If
these trends match the trends expected of a parallel cylinder capacitor then it is likely that this is an
accurate model.
Additionally, the current dielectric is not a pure dielectric but more of a strong resistor. Modifying the
dielectric material should significantly change the capacitance.
Inductance
Similar to the capacitance test, adjusting the physical dimensions (diameter, thickness, and length) of
the sample should create trends in the data. Changing the dielectric should also accentuate these
effects.

41

3.4.Sensor Skin Development


3.4.1. Requirements
Oil and Gas Marine Hose
Attribute
Curing Temp
Operation Temp (oil, gas)

Range
300-320F for 1 hour
40F to 140F Product Temperature
-40F Interior Temperature
-4F to 180F
-65F to 300F
6x ID (~10% in simplified model, see below)
6 to 40 inches
2-5 years
Water, oil, gas
Abrasion, weather, and UV resistant
Primary Carcass
Visible, audible, data transmission (currently diver inspection)
$30k-40k USD

Operation Temp (hydraulic)


Bend Radius
Product Size (ID)
Product Life (oil,gas)
Chemical Resistance
Surface Resilience
Leak Detection
Damage Communication
Product Cost (oil,gas)

Calculations for 10% strain

Ro
Ri
R

ID
Neutral Axis (NA)

Given the relationships above, the strain on the


inside and outside of the ring can be calcualted. Assuing a straight section of hose that is bent to a
radius that is 6 times the inner diameter (shown above as a complete circle for demonstration
purposes), the inner circumferance and outer circumferance were both intially the same length as the
neutral axis. The change in length (strain) can therefore be calculated by comparing the circumferance
of the neutral axis to the inner and outer circumfrance of the ring.
Compression
42

Given strain:

Where
(

And, based on the definitions above

[(

Substituting back into the strain equation,

Tension
Given strain:

Where
(

And, based on the definitions above

[(

Substituting back into the strain equation,

43

Therefore, the strain expected on the sensor is


. For simplicity and to accomidate for some of
the assumptions made in the model above, the strain requirement will be assumed to be
3.4.2. Form Factor
Several form factors were considered for the design of the sensor skin and their advantages and
disadvantages weighed as shown below.
Advantages

Disadvantages
Continuous Sensor Skin

Simple Concept
Difficult to apply to hose
Many material options already available
Requires adhesive
Easy to manufacture
Promising materials are expensive
Strip Tape Sensor Skin
Easy to apply
Difficult to control quality of application
Many material options already available
Requires adhesive
Easy to manufacture
Promising materials are expensive
Risk of leakage between tape layers
Paint on Sensor Skin
Very easy to apply
Chemical tuning required
Easy to manufacture
Extensive testing required
Novel idea
Easily scalable
Initially, a strip tape sensor seemed like the best option, but a potentially disastrous failure mode was
conceived, shown below, that led to further development of the paint on sensor skin. In the event of
continued pressure against the tape, water or corrosive fluid may deteriorate the bond between the
tape and the hose and the tape on itself allowing fluid to leak through without shorting the sensor skin.

Bonded Tape

Over Time

Debonded Tape

Pressurized Fluid

Pressurized Fluid
Figure 23: Debonding Tape Illustration

44

3.4.3. Electrode Design


Once the paint on form factor was chosen, it was clear that a custom material would need to be
developed for the paint on sensor skin application. During this process several options were considered
for the electrode, dielectric, and backing material.
Initial Design Options
Electrode
Super Shield Coating
Carbon-doped Rubber Cement

Dielectric
Pure Rubber Spray
Rubber Cement

Carbon-doped Liquid Tape

Liquid Tape

Backing
Paper
Adhesive Functionalized
Neoprene
Soft Rubber

Super Shield Coating with Pure Rubber Spray on Paper

Figure 24: Super Shield and Rubber Spray Coating images courtesy of MG Chemical and Blue Magic respectively

The Super Shield Coating with Pure Rubber Spray on paper was assembled per the assembly procedure
which can be found in the appendix from January 22, 2012 (arranged chronologically). Key elements are
summarized here with representative photographs.

45

Figure 25: Completed Sensors partially separated

Surface resistance test


Manufacturers Data

Figure 26: Surface resistivity vs thickness plot of Super Shield courtesy of MG Chemicals
http://www.mgchemicals.com/products/841.html accessed 3/21/2012

Experimental Data
46

Surface resistance tests were conducted using a four probe test method shown below.

Figure 28: Schematic of 4


probe test courtesy of
Vessels42 of the Wikimedia
Commons

Figure 27:4 Probe test on Super Shield


Sample

A four probe test is the industry standard for measuring sheet resistance. The reason for this is that the
four probe tests applies a constant current from a precision current source (in this case a Keithly 6220).
By modeling the surface of the material as a resistor, Ohms law can be used to demonstrate that a
voltage drop will be produced between any two points on the surface between the probes providing the
current.

Where

is the voltage drop, is the constant current, and

is the resistance in the sheet. This means

that there will be a voltage drop across probes 2 and 3 (as indicated in Figure 32). This voltage drop can
be measured using a voltmeter.
The advantage to the four probe test actually lies in its ability to negate the effects contact resistance.
Since a voltmeter utilizes an extremely high internal resistance in the meter to make measurements,
virtually no current passes through the meter and therefore virtually no current passes across the
contacts. Again, referring to Ohms law, it can be shown that if as the current across a resistance
approaches zero, the voltage drop across that resistance will also approach zero.

47

A four probe resistance test was run on each Super Shield sample (SN 013 through 024) with the
following results:
Note: Sample 013, 017, 020, 022 have been omitted as extreme outliers.

Four Probe Data, 0.2" Spacing


20
Sample 014
Sample 015
Sample 016
Sample 018
Sample 019
Sample 021
Sample 023
Sample 024

15

Voltage (mV)

10
5
0
-5
-10
-15
-20
-10

-8

-6

-4

-2

0
2
Current (mA)

10

Figure 29: Four Probe data for Super Shield Spray on electrode

In the figure above, note the linearity of each of the samples. Based on Ohms law, the slope of each line
represents the resistance of each sample. The resistance ranges from about 0.7 to about 1.7 . This
indicates that the sample is highly conductive and electrically well suited for the application.
Thermal Testing
Qualitative thermal testing was done to the Super Shield and Pure Rubber spray sample by pouring
liquid nitrogen over the sample until it was evenly cooled to less than -60 C. The temperature
48

distribution of the sample was recorded using a FLIR T640 (SN: 55901055) thermal imaging camera. The
visible light and infrared images are below. The metallic electrode on the sample appears to be warmer
than the rest of the sample as a result of reflecting infrared waves from the background off the nickel
particles on the samples. The sample is approximately 0.020 0.010 thick so it would be evenly
cooled by pouring liquid nitrogen over the sample. The distinct outline of the metallic section is another
indication that the apparent temperature variation is actually reflection.

Figure 30: Cooling a Super Shield on paper backing sample with liquid nitrogen

Figure 31: Bending a Super Shield on paper sample at -60C

49

These thermal images show that the material was able to survive extreme cold (-60 C) without reaching
a glass transition temperature (the temperature at which a polymer becomes extremely brittle, like
glass). The material was manually bent and folded without visible cracking. Micro-cracking may have
occurred.
Failure Testing
Testing was conducted to determine if the Super Shield electrode and pure rubber spray sensor would
indicate a failure in the event of a puncture. For this, the resistance was measured between the two
electrodes of an undamaged sensor sample as shown in Figure 36. The sensor was then pressed very
firmly against with the blunt end of a sharpie. This action did not produce a measureable change in
resistance between the electrodes. The sensor was then punctured with a steel razor blade and the
resistance changed. The razor blade was then removed to ensure that the connection was not the result
of the conductive steel. After the blade was removed, the sensor was still shorted as shown in Figure
37.

Figure 32: Resistance between electrodes before puncture

Figure 33: Resistance between electrodes after puncture


had occurred and puncturing item was removed

These tests confirmed that, electrically, the Super Shield electrodes and pure rubber spray would work
for the desired application.
50

Strain Testing
As discussed in 3.4.1, the sensor must be able to survive strain cycles of 10%. This test was conducted
by applying the spray on sensor to a soft neoprene rubber that was 1/8 inch thick. The sensor was
applied to the side of the neoprene that was functionalized for an adhesive. The sensor was constructed
on February 15, 2012. The fully assembly procedure can be found in the appendix. Some issues were
discovered when the masking tape was removed from the sample. It turned out that the adhesion
between the masking tape and the dielectric was stronger than the adhesion between the electrode and
the functionalized neoprene. This was worked around by not removing all of the tape and the test was
continued. The samples also had copper leads bonded to the front and back electrodes using
conductive epoxy to provide something for the alligator clips to clip to during the strain test.
Once the samples were created, each sample was loaded into the Instron tensile test machine and
slowly strained to 10% tensile strain. The copper leads were used as the probes of a four probe test.
The voltage was measured across the inner probes and data automatically recorded as the sample was
strained. The data is provided below

Dynamic 4 Probe Test, 0.3 Total Extension at 3/min, sample 030

51

Stop
Start

Figure 34: Dynamic 4 probe test of a sample that performed well (did not crack)

Dynamic 4 Probe Test, 0.3 Total Extension at 3/min, sample 033

Stop
Start

Figure 35: Dynamic 4 probe test of a sample that that cracked. This was typical of the super shield spray on electrode

The data shows an increase in resistance, often to the reading limit of the test equipment. This is most
likely the result of the electrode cracking during the strain process. The following image was taken after
only one strain cycle. This type of cracking was typical of the Super Shield spray on electrode samples.

52

Figure 36: Super Shield sample strained to 10% showing severe cracking

These results indicated that this sensor design would not work well for the desired application. It was
determined that a new electrode material must be developed for the application.
Custom Electrode Development

Three factors were considered when designing electrode material: material matrix, graphite content,
and strain. Electrodes were formed with a polymer matrix material of liquid paint or rubber cement.
They contained a graphite content of 50, 100, 150, 200, and 250 percent by volume, and strain values of
0 to 10 percent in tension. Three samples were taken at each treatment level and an analysis of
variance (ANOVA) was performed on the collected data to determine the effect of different treatment
and the resistance of the electrodes. Optimal results will have a minimal resistance and inversely have a
high conductance.
Strain Testing

53

The electrodes were formed on a flexible neoprene backing so that they could be loaded into a tensile
test machine. Four probes were connected linearly along the electrode. The outer two probes were
connected to a precision constant current source, while the inner probes measured the voltage drop
across a known distance. Resistance could then be calculated using Ohms law

. The resistance

of the electrodes for the various treatment levels can be found in the appendix. Please see the appendix
for a detailed test procedure. The test was run on February 24, 2012.
Some measurements were outside the reading limits of the test instrument. When this was the case,
the highest possible reading for the test instrument was assumed. Once the database was finished,
code was written to import the database into MATLAB and run a three factor analysis of variance
(ANOVA).

Figure 37: Photograph of experimental setup

54

Figure 38: Liquid Paint matrix with 250% graphite by volume strained at 10%

Figure 39: Rubber Cement matrix with 250% graphite by volume strained at 10%

Results
'Source'

'Sum Sq.'

'd.f.'

'Singular?'

'Mean Sq.'

55

'F'

'Prob>F'

'A'

[1.8208e+011]

1]

0]

[1.8208e+011]

[ 38.1102]

[1.8880e-008]

'B'

[1.4088e+012]

1]

0]

[1.4088e+012]

[294.8684]

[3.8184e-030]

'C'

[2.6549e+011]

4]

0]

[6.6371e+010]

[ 13.8916]

[7.2894e-009]

'A*B'

[3.5297e+010]

1]

0]

[3.5297e+010]

7.3877]

0.0079]

'A*C'

[6.6018e+010]

4]

0]

[1.6505e+010]

3.4544]

0.0113]

'B*C'

[7.5012e+009]

4]

0]

[1.8753e+009]

0.3925]

0.8135]

'Error'

[4.3000e+011]

90]

0]

[4.7778e+009]

[]

[]

'Total'

[2.5114e+012]

[ 105]

0]

[]

[]

[]

From the results it is important to note the p-values for all of the factors are well below an accepted
alpha value of 0.05, except for the interaction between the matrix material and the graphite
concentration. This indicates that the all factors and interactions are statistically significant in
determining the electrical conductivity, except for the interaction between the matrix material and the
graphite concentration (B*C).

When observing Figure 40, note the peak at 100% by volume graphite content indicating a highest
resistance value. The lowest resistance value (most desirable result) is found at 200% graphite content
by volume. There is also strong interaction between strain and matrix material as well as strain and
graphite content. This interaction suggests that further work should be done examining the effects of
strain on the matrix material and strain on graphite content individually.

56

x 10

mean

2.5

1.5

10
Strain

1
2
Matrix Material

50 100 150 200 250


Graphite Content (% by volume)

Figure 40: Main Effects Plot (Matrix Material: 1-Liquid Tape, 2-Rubber Cement

When observing Figure 41, the same inferences can be made as the ANOVA table and the main effects
plot (Figure 40), however this plot breaks down the results even further for inspection.

57

50 100 150 200 250


4
A =0
A = 10

2
1
5

5
xx 10
10

x 10
4

B=1

B=2

1
5

x 10
4

C = 50

C = 100

C = 150
C = 200
C = 250

10

Figure 41: Interaction Plot. A: Strain %, B: Matrix Material (1-Liquid Tape, 2-Rubber Cement), C: Graphite Content % by
volume

58

Normal Probability Plot


0.997
0.99
0.98
0.95
0.90

Probability

0.75

0.50

0.25

0.10
0.05
0.02
0.01
0.003
-0.3

-0.2

-0.1

0.1

0.2

0.3

0.4

Data

Figure 42: Normal Plot of Residuals

Figure 42 shows a normal distribution plot of the residuals. It can be seen that the residuals of the data
collected do not follow a normal distribution. Once the significant factors were determined a linear
regression was performed on the data set taking into account the significant factors and significant
interaction. The regression can be seen in the equation below.
( )

)]

Conclusion
These results were expected. Graphite, a conductive material, was placed in an insulating matrix so
increasing the graphite content should also increase the conductivity. However, because the graphite
particles are discrete, adding too much graphite may produce discontinuities in the conductive path

59

when strained. This is evident in a reduced conductance at 250% graphite by volume as compared to a
200% graphite by volume.
The liquid paint used in the experiment was a liquid, paint-on electrical tape. This matrix was a bit
thicker than the rubber cement and is also designed to be a strong dielectric. These two properties
likely contributed to the poorer conductivity in the sampled using the liquid paint matrix as compared to
the rubber cement matrix.
Therefore the optimal electrode for conductivity and surviving strain is around 200% graphite content
and has rubber cement as its binding matrix.
The change in resistance of the electrode composite created could be used as a strain gauge. If the
electrode is attached to a surface the strain of the surface could be measured by recording the change in
resistance, assuming that the electrode composite does not affect the elasticity of the system.
As for the residuals not following a normal distribution, the most likely culprit is poor mixing of the
graphite into the matrix during fabrication. Two possible solutions to this problem are 1) using an
ultrasonic mixer or 2) use a shear plane mixer to mix the composite. The ultrasonic mixer uses high
frequency sounds waves to shake up the mixture. The shear plane mixer uses two plates that are very
close together to create a shear-plane (think boundary conditions in fluids) stirring of the mixture.
Another issue is testing if the composite is uniformly mixed. To achieve enough resolution to observe
the uniformity of the composite a high-powered microscope (SEM for example) would be needed.
Unfortunately the SEM requires the sample to be under vacuum. Putting the composite in question in a
vacuum may cause out gassing, which could damage the microscope and affect the results.
Rubber Compatibility

60

The electrode paint has two functions: behave as an electrode and act as an adhesive to bond the
dielectric paint to the hose. For an adhesive to be effective, it must wet the substrate, not dissolve the
substrate, and harden. A simple test was performed where the mixture of 1:2:2 parts rubber cement to
graphite to toluene by volume was created and dripped onto various rubber samples. Photographs
were taken level with the sample so that the contact angle of the adhesive and the rubber substrate
could be observed and measured. A smaller contact angle indicates better wettability and therefore an
increased likelihood that the adhesive will bond well to the rubber substrate.
Rubber Sample
Lot Number
3192R4 (EPD)
3187R1 (EPD)
3169R4 (EPD)
3176R6 (EPD)
3179R3 (EPD)
3155R1 (EPD)
D1124H (HPD)
A508H (HPD)
D620 (HPD)

Contact Angle
D1124H
40
45
50
40
40
75
40
50
50

D620

Figure 43: Contact angle of RC200 electrode paint on various rubber


samples supplied by Parker Hannifin

3.4.4. Custom Dielectric Development


After the rubber cement-based 200% by volume graphite content electrode (RC200) was chosen, it was
used to make several sensor skin samples. Unfortunately, all of these samples were unusable because
the electrodes were not electrically isolated. After some thought, it was determined that the toluene in
the electrode was dissolving the Blue Magic Pure rubber spray. A simple test was designed to test this
hypothesis and whether latex (known for chemical stability in gloves) would hold up against toluene: six
small Petri dishes were filled half way with rubber (3 spray rubber, 3 liquid latex) and allowed to dry

61

entirely in a fume hood. The dishes were then filled the rest of the way with toluene and covered. The
dishes were checked at 5 minutes, 15 minutes, and 30 minutes to determine if the toluene was
dissolving the rubber. After 30 minutes, the Blue Magic rubber spray had completely dissolved while
the latex was approximately unaffected.

Figure 45: Blue Magic pure rubber spray completely


dissolved after 30 minutes soaking in toluene.

Figure 44: Ben Nye pure latex rubber unaffected after 30


minutes soaking in toluene.

Sensor Skin Samples with Latex Dielectric


The results of this simple test prompted the use of liquid latex as a dielectric. Additional sensor skin

samples were created on April 3, 2012. These samples, painted on 1/8" neoprene that was
functionalized for adhesive by the manufacturer, were made by painting a layer of RC200 electrode
paint followed by a layer of latex and then an additional electrode layer. To test how many coats of
latex were necessary, 3 samples were made of each set: 1 coat, 2 coats, 3 coats, and 4 coats. Each layer
was painted using a small paint brush. Care was taken to ensure that the electrodes did not make
contact. Each layer was allowed to dry before the next layer was applied.

62

Once the samples had fully cured, tests were run to determine the change in resistance between the
two electrodes. The resistance was recorded before damage was induced and then damage was
induced by piercing the sample with a razor blade. The resistance consistently dropped several orders
of magnitude upon damage being induced by the razor blade. To minimize the conductive effects of the
razor blade, the blade was coated in epoxy. However, inspection under a microscope indicated that the
edge of the blade was not fully coated. To account for this the blade was inserted and removed before
measurements were recorded. The resistance measurements were very similar with the razor blade in
the sample and removed so it had little effect.

Figure 46: Representative sample of RC200 and latex


sensor before damage was induced.

Figure 47: Representative sample of RC200 and latex


sample after damage was induced by piercing the sample
with a razor blade and removing the razor blade.

To simulate a burst failure, samples were torn in the plane of the sensor. To simulate foreign object
damage that would cause the hose to split with one layer effectively going below the other, the sensor
was torn perpendicular to the plane of the sensor. Each of these tests were run by, first, measuring the
initial resistance between the electrodes and then scoring the back of the sensor (about half way
through the backing) in a line starting from the middle of the sensor and moving outward toward an
edge. The sample was then gripped on either side of the scoring and torn with pliers in the appropriate
direction.

63

Figure 49: Schematic of in plane tear failure

Figure 48: Schematic of out of plane tear


failure.

The results of each of both of these tests were very similar and simple: the latex dielectric stretched
rather than tearing so the electrodes never made contact. Typically, no change was observed in the
resistance between the electrodes as a result of this induced failure. Please see the table of data below.

Figure 50: Sensor with an in plane tear failure.

Figure 51: Sensor with an out of plane tear failure.

SN

Electrode

Dielectric

Coats of
Dielectric

112
113
114
115
116
117
118
119
120
121
122
123

RC200
RC200
RC200
RC200
RC200
RC200
RC200
RC200
RC200
RC200
RC200
RC200

Latex Paint
Latex Paint
Latex Paint
Latex Paint
Latex Paint
Latex Paint
Latex Paint
Latex Paint
Latex Paint
Latex Paint
Latex Paint
Latex Paint

1
2
3
4
1
2
3
4
1
2
3
4

64

Initial
Resistance
(kOhm)
inf
300
inf
inf
inf
50
inf
inf
inf
inf
inf
inf

Failure Mode
Pinhole
Pinhole
Pinhole
Pinhole
In Plane Tear
In Plane Tear
In Plane Tear
In Plane Tear
Out of Plane Tear
Out of Plane Tear
Out of Plane Tear
Out of Plane Tear

Final
Resistance
(kOhm)
14
10
22
100
inf
150
inf
inf
inf
inf
inf
inf

Upon realizing that the latex dielectric was actually too flexible,
an effort was made the tune the dielectric. To do this, latex
was mixed with different powders in an effort to stiffen and
weaken it to behave more like the electrode material.
The following powders were chosen to be added to the liquid
latex:

Walgreens Body Powder


Salt
HX-CECO Latex Thickener

Figure 52: Unstrainted Latex Sample loaded in


Instron Machine

Each of these was added at the following volume


percentages*:

5%
10%
20%
40%

*Mass percent would be a more accurate measurement, however, this was accidently
overlooked until Ms. Anne Brant pointed it out after the assembly has been completed.

Finally, two new paints were introduced:

Ace Hardware Latex Paint (traditional latex paint)


Ace Hardware Alkatex Paint (latex paint designed to
behave similar to oil paint)

For most of these combinations (some combinations separated


the latex and clumped such that it was impossible to paint on)
three sensor skin samples were made to test at least one of the
failure modes on each. The material was also cast into
rectangular Teflon molds so that a stress strain curve could be
generated as the material was pulled to failure.

Figure 53: Strained Latex Sample loaded in


Instron Machine

65

Even with the thickeners and additives, the


latex was far too flexible for the intended
dielectric application. The latex was so
flexible that, in most cases, it deformed
and slipped from the clamps rather than
tore. However, the electrode material was
also tested and found to be quite weak
and brittle. Most of the material that was
poured into the mold evaporated off
leaving only a thin sample to be tested.
This sample, however, readily cracked and

Figure 54: RC200 Electrode Sample loaded, unstrained, in the


Instron 5984

flaked off and therefore needed to be


handled with great care and tested on the more sensitive Instron 5984 (SN: 5948R7176) with a 5N max
static load cell (SN: 103583). A representative selection of plots from the tensile tests is included below.

66

RC200 Electrode
0.35
0.3
0.25

Stress (MPa)

0.2
0.15
0.1
0.05
0
-0.05

0.05

0.1

0.15

0.2

0.25

Strain

Figure 55: Stress Strain plot of RC200 Custom Electrode Materail

Latex, Body Powder X%, Rate: 5in/min


0.9
SN: 164, 5%
SN: 167, 5%
SN: 168, 10%
SN: 169, 10%
SN: 170, 20%
SN:172, 20%
SN: 173, 40%
SN: 175, 40%

0.8
0.7

Stress (MPa)

0.6
0.5
0.4
0.3
0.2
0.1
0

3
Strain

Figure 56: Stress Strain plot of Latex mixed with Body powder at varying concentrations

67

Latex, HX-CECO Thickener X%, Rate: 5in/min


1
SN: 215, 5%
SN: 216, 5%
SN: 218, 10%
SN: 219, 10%
SN: 221, 20%
SN:222, 20%
SN: 224, 40%
SN: 225, 40%

0.9
0.8

Stress (MPa)

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0

4
Strain

Figure 57: Stress Strain plot of Latex mixed with HX-CECO thickener at varying concentrations

The most important aspect of the plots above is the strain to failure. Ideally, the sensor will fail shortly
after the hose exceeds its minimum bending radius. In this case, that is a strain of about 10%. For the
sensor to work, the electrodes must survive beyond that and, based on the above plot, it does. In fact,
the electrode survives to approximately 20%. However, the dielectric must fail before the electrode. As
shown in Figure 55, if the dielectric is too flexible it will not allow the electrodes to make contact.
This exact issue is evident in the stress strain plots of the latex mixtures. In every experiment the
sample survived strains well over an order of magnitude larger than is desired. Furthermore, the
failure points, in most cases, are actually only the points at which the sample slipped from the clamp.
In fact, only 4 samples (serial numbers 215, 217, 219, 220) failed. And those failed very close to the
clamp jaw suggesting that stress concentrations were a greater factor than the material property. In
hindsight, a dog bone shape sample would have been preferable over a rectangular sample to reduce
the impact of stress concentrations at the clamp jaw.
68

It is also interesting to note that the samples did get progressively stiffer as the
fraction of additive increased. The body powder has a relatively minimal and
inconsistent effect as compared with the commercial Holdens HX-CECO latex
stiffener. The HX-CECO stiffener behaved very predictably to produce the
desired effect, but not to the desired degree. Upon inspection of the material, it
was determined that adding the amount of stiffener necessary to achieve the
desired strain to failure would have stiffened the mixture into a very thick putty
rather than a paint and therefore would not be suitable for the application.

Figure 58: Paint


brush with cured
latex stuck to it as a
result of premature,
salt-induced curing

Salt was also used, but mixing salt with the latex very quickly caused the latex to clump and cure. It
became completely unusable before the test could be conducted.
Despite unfavorable stress strain results, these dielectric materials were used to construct sensors which
were then tested using the Manual Sensor Failure test procedure (4/20/2012). This procedure calls for a
pinhole test where the sensor is punctured with a sharp, pointed object to pierce the sensor, simulating
a pinhole failure and tearing out of plane and in plane as shown in Figure 54 and Figure 55 respectively.
Tempera
Efforts made to tune the dielectric were not successful enough to produce a working sensor. Ms. Anne
Brant suggested the use of several materials such as craft tempera paint. The specific tempera paint
used was orange Aurora Tempera Paint. Orange was chosen to sharply contrast with the black and dark
grey backing material and electrode material.
5/16/2012 The first test compared the behavior or a diluted tempera mixture of 2:1 tempera to water to
a mixture of 1:1:2 of Ben Nye Latex, Holdens HX-CECO latex thickener and water, respectively. Each
mixture was sprayed onto a piece of foil using a Badger Airbrush with a fully open nozzle and air
69

pressure set to 30PSI. Approximately 4-5 steady passes were made with the air brush over a simple
rectangular stencil for each sample. The thicknesses are given in the table below in thousandths of an
inch.

Sample
Foil
1 Tempora
2 Tempora
3 Tempora
4 Latex (1:1)
5 Latex (1:1)
6 Latex (1:1)

1
1.10
1.45
1.50
1.35
2.20
3.30
2.60

2
0.90
1.45
1.25
1.25
2.20
2.80
2.55

3
0.90
1.35
1.40
1.40
2.80
2.05
2.30

4
0.80
1.30
1.30
1.30
2.00
3.30
1.50

5
0.85
1.55
1.25
1.25
2.10
2.70
2.40

Std.
Average Dev
0.91
0.11
1.42
0.10
1.34
0.11
1.31
0.07
2.26
0.31
2.83
0.52
2.27
0.45

Tempera Samples

Latex Samples

Figure 59: Tempera and Latex samples on foil

70

Coating Electrically
Thickness Insulated
NA
Yes
0.51
Yes
0.43
Yes
0.40
Yes
1.35
No
1.92
Yes
1.36
Yes

After the electrical isolation tests was run, a tear test was run to qualitatively compare the two
dielectrics in terms of how readily they tore when the foil tore. Even with a very thin application and
very sharp tear from the foil, the latex stretched across the tear such that electrodes on either side
would not be able to make contact. The tempera, however, tore readily and appeared to be quite
brittle.

Figure 61: Latex sample stretching across tear during foil


tear test

Figure 60: Tempera sample showing clean tear on foil


tear test

5/17/2012 Another round of tests was run with another attempt to tune the latex and test application
methods. The idea, this time, was that a very thin coating would help the latex tear more readily. To
achieve this, a mixture of 1:2:4 of Ben Nye Latex, Holdens HX-CECO thickener, and water, respectively,
was created. Three samples were airbrushed onto foil and three samples were paint brushed onto foil.
The mixture sprayed very thin so the foil was sprayed continuously for one minute, moving
continuously, over each 1 x 1.5 stencil with the air at 30 PSI. The brush-painted samples did not wet
the aluminum well and resulted in some beading. After drying, the airbrushed samples were not
electrically insulating and the brushed samples did not tear with the foil. The thickness and electrical

71

insulation data for these samples are in the table below. Thicknesses are given in thousandths of an
inch.

Sample
Foil
1 Airbrush
2 Airbrush
3 Airbrush
4 Paintbrush
5 Paintbrush
6 Paintbrush

1
0.90
0.90
0.85
0.75
2.60
3.15
2.80

2
0.80
0.85
0.75
0.85
2.20
3.00
1.70

3
0.85
0.75
0.80
0.80
3.05
1.95
2.35

4
0.75
0.90
0.85
0.90
2.30
1.80
1.50

5
0.80
0.80
0.85
0.80
2.00
1.90
2.60

72

Average
0.82
0.84
0.82
0.82
2.43
2.36
2.19

Std.
Dev
0.06
0.07
0.04
0.06
0.41
0.66
0.57

Coating
Thickness
NA
0.02
0.00
0.00
1.61
1.54
1.37

Electrically
Insulated
NA
0.3
0.3
2.4
Yes
Yes
Yes

Airbrushed

Paintbrushed

Figure 63: Brushed, thickened latex


stretching over a tear in foil

Figure 62: Thickened latex samples. 1-3 Airbrushed, 4-6 Paintbrushed on to foil

5/18/2012 The next step in the progression was a mix of tempera and latex. This mixture was 1:1:2 of
Ben Nye latex, Aurora tempera paint, and water respectively. This mixture sprayed well in the Badger
Airbrush. Each of the first three 1 x 1.5 stencil samples was sprayed continuously for approximately 7
seconds with the air pressure at 30 PSI. The mixture was applied with a paintbrush for the next three
samples. The paintbrush application had difficulty wetting the foil. Thicknesses given in the table below
are in thousandths of an inch.

Sample
Foil
1 Airbrush
2 Airbrush
3 Airbrush
4 Paint Brush
5 Paint Brush
6 Paint Brush

1
1.10
1.45
1.50
1.35
2.20
3.30
2.60

2
0.90
1.45
1.25
1.25
2.20
2.80
2.55

3
0.90
1.35
1.40
1.40
2.80
2.05
2.30

4
0.80
1.30
1.30
1.30
2.00
3.30
1.50

73

5
0.85
1.55
1.25
1.25
2.10
2.70
2.40

Std.
Average Dev
0.91
0.11
1.42
0.10
1.34
0.11
1.31
0.07
2.26
0.31
2.83
0.52
2.27
0.45

Coating Electrically
Thickness Insulated
NA
NA
0.51
Yes
0.43
Yes
0.40
Yes
1.35
No
1.92
Yes
1.36
Yes

Airbrushed

Paintbrushed

Figure 64: Latex and tempera samples airbrushed and paintbrushed onto foil

Figure 65: Latex and tempera sample stretching across a tear in aluminum foil

5/30/2012 The above tests showed tempera as a very plausible dielectric material while latex was
shown to be far too elastic, even when additives and mixtures were used. However, based on the foil
tear test, there was a concern that the tempera may be too brittle for use in the sensor skin application.
To test this, pure tempera and a 3:1:2 tempera: latex: water mixture were each sprayed onto adhesive
functionalized neoprene rubber. For each sample, a 3 x 1 piece of rubber was used and was sprayed
with the mixture for approximately 15 seconds with a Badger airbrush with the air pressure at 30 PSI.
Each sample was then loaded into a vice and brought to a neutral point. Tension was then applied until
74

cracks in the material were visible under a microscope. Data from the samples is recorded in the tables
below. Thicknesses in the table below are in inches.

Tempera : Water
Sample #
300
301
302
303
304
305
306
307
Rubber

1
0.11765
0.11305
0.11165
0.11757
0.1155
0.1109
0.1121
0.1123
0.11045

2
0.11775
0.11405
0.11015
0.1119
0.11615
0.11055
0.11515
0.11075
0.10985

3
0.12015
0.1125
0.1185
0.1136
0.118
0.114
0.1118
0.1134
0.112

4
0.11685
0.1125
0.11825
0.1182
0.11205
0.11205
0.1195
0.11605
0.11095

Standard
Coating Initial Final
Strain
5
Deviation Average Thickness Length Length Strain Tear
0.11685 0.00135 0.11785 0.00699 1.875 2.875
0.53
0.11175 0.00085 0.11277 0.00191
1.75
2.875
0.64
Yes
0.11825 0.00411 0.11536 0.0045
1.75
2.25
0.29
Yes
0.11815 0.00293 0.115884 0.005024
1.5
1.875
0.25
Yes
0.1121 0.00262 0.11476 0.0039
1.5
1.75
0.17
Yes
0.1189 0.00342 0.11328 0.00242 1.375
1.75
0.27
Yes
0.1118 0.00335 0.11407 0.00321 1.625 2.125
0.31
Yes
0.1123 0.00197 0.11296 0.0021
1.375 1.875
0.36
Yes
0.11105 0.00080 0.11086
NA
NA
NA
NA
NA

Cut
Tear
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
NA

3
0.10885
0.10845
0.1119
0.1118
0.11085
0.1093
0.1113

4
0.1094
0.1121
0.1137
0.1103
0.1096
0.1108
0.11035

Standard
5
Deviation Average
0.10825 0.00043 0.10894
0.1089 0.00152 0.11003
0.111
0.00183
0.1111
0.11005 0.00071 0.11055
0.10955 0.00110 0.10929
0.11095 0.00108 0.10975
0.11065 0.00073 0.11123

Cut
Tear
Yes
Yes
Yes
Yes
Yes
Yes
NA

Tempera : Latex : Water


Sample #
308
309
310
311
312
313
Rubber

1
0.1091
0.10965
0.10895
0.1104
0.10815
0.10925
0.11205

2
0.1091
0.11105
0.10995
0.1102
0.1083
0.10845
0.1118

Coating Initial
Thickness Length
-0.00229 1.125
-0.0012
1.125
-0.00013 1.125
-0.00068
1.25
-0.00194 1.125
-0.00148 1.125
NA
NA

Final
Strain
Length Strain Tear
3
1.67
No
3
1.67
No
3
1.67
No
3
1.40
No
3
1.67
No
3
1.67
No
NA
NA
NA

Please note that the standard deviation of the thickness measurements is approximately the same magnitude as the coating
thickness (which was calculated as the difference between the average rubber thickness and the average sample thickness with the
coating. Therefore, coating thickness values only show that the coating was very thin and gives little indication of actual thickness.
Obviously it is impossible to have a coating that is a negative thickness.

Figure 67: Sample loaded into vice and brought to a


strained state

Figure 66: Cut tear sample showing that the coating does
not stretch across the tear

75

In response to the promising results obtained from the tempera and


the latex and tempera mixture sensors were constructed using the
RC200 electrode paint and a dielectric of tempera. The first round of
tests, performed on 6/6/2012, showed that a single coating of
tempera sprayed onto the samples for approximately 3 minutes at
20PSI air pressure failed to insulate the paint brushed electrodes.
None of the six samples tested were electrically insulated (<15M).
These results are in conflict with the tests performed on foil. The foil
tests suggested that a single coating of tempera paint would
sufficiently insulate the electrodes. The most significant difference
between the foil and the paint brushed electrode was surface

Figure 68: Representative sensor


sample with RC200 electrode material
and one coat of tempera paint as a
dielectric.

roughness. The paint brushed electrode is much rougher than the smooth foil which allows for crevices
that were not fully coated by the tempera dielectric.
6/12/2012
To remedy this problem, several sensors were constructed with varying thicknesses of dielectric (2
through 5 coats) to determine how many coats were necessary to insulate the electrodes. Each coat
was sprayed on until there was a sheen to the paint. Care was taken to ensure that no dripping
occurred and the samples were evenly coated. This took a bit of care because the samples had to be set
up vertically for the airbrush to work properly. (The airbrush must be used with the stick part of the
airbrush held approximately parallel to the ground so the target should be approximately vertical.) The
electrical insulation of each sensor was tested using a Fluke 16 multimeter. The samples were
considered isolated if the resistance was too high for the multimeter to read (>15M).

76

2 Coats

3 Coats

4 Coats

5 Coats

Figure 69: Sensor skin samples with varying thicknesses of orange tempera dielectric paint. From left to right there are 2, 3,
4, and 5 coats

2 Coats

3 Coats

4 Coats

5 Coats

Figure 70: Microscope images of the different coating thicknesses

The microscope images clearly show an increasing dielectric layer thickness. The most notable change is
between 3 coats and 4 coats. With only 2 or 3 coats the surface roughness of the electrode material is
still very evident and graphite can be seen through the translucent tempera paint. At 4 coats the
surface roughness begins to smooth out and the graphite cannot be seen through the paint. At five
coats it is clear that the sensor is thoroughly covered and that there are no longer gaps or crevices
exposed for the second layer of electrode material to seep into.

77

Sensor
Number
320
321
322
323
324
325
326
327
328
329
330
331

# of Coats
2
2
2
3
3
3
4
4
4
5
5
5

Insulated?
no
no
no
no
no
no
yes
no
no
yes
yes
yes

Failure Mode
NA
NA
NA
NA
NA
NA
Pinhole
NA
NA
Pinhole
Out of Plane Tear
In Plane Tear

Final
Resistance
NA
NA
NA
NA
NA
NA
2.3 k
NA
NA
4.0 k
85 k
.24 k

While the five coat samples were very successful in the previous test, there were only three samples. A
second test with 16 samples was run to determine the repeatability of those results. For these samples,
the same procedure was followed with the same materials (RC200 electrode material and 2:1 tempera
to water on adhesive functionalized neoprene). To standardize the coat application, each coat was
sprayed continuously for 5 minutes at 20PSI air pressure to the entire batch of 16 samples. Each coat
was allowed to dry fully before the next coat was applied. 4 coats of the RC200 conductive paint were
applied by paint brush as the electrodes.
Once the sensors had finished curing, a Fluke 16 multimeter was used to check the resistance between
the two electrodes. A sensor was considered insulated if the resistance between the electrodes was
greater than 15M. Of the 16 samples constructed, only 7 were considered insulated. This suggests a
need for improved process control and a new construction process that is less sensitive to variability.

78

Not Insulated

Insulated
Figure 71: Several sensor skin samples with 5 coats of dielectric each. The samples are separated into insulated and not
insulated as some of the dielectrics were not sufficient to insulate the electrodes.

Only the insulated samples were allowed to continue for failure mode testing. The data from the failure
mode testing, shown below, was very promising. Every sensor responded well to the failure mode it
was subjected to. The resistance between the electrodes dropped at least 4 orders of magnitude in
each case.
Sensor
#
333
334
335
338
342
345
346

Failure Mode
Pinhole
Tear out of Plane
Tear In Plane
Pinhole
Tear Out of Plane
Tear In Plane
Pinhole

Final
Resistance
5.7 k
1000 k
7.84 k
2.3 k
101 k
45 k
6.0 k

Figure 72: Microscope image of 5 coat


tempera sensor skin sample

However, the bad samples (those which were not electrically insulated) did not go to waste. The
flexibility of the tempera dielectric was tested using a very thin layer. The tests have shown that a much
thicker layer was necessary to insulate the two rough electrodes. The bad samples were used to
quickly test the flexibility of this new sensor configuration. One sample was loaded into the vice and
79

strained while the response to the strain was observed both at a microscopic and macroscopic level.
The tempera cracked readily with moderate bending as the sample was loaded into the vice and
continued to crack as it was strained.
6/18/2012
To test if the dielectric was still flexible when applied in a thicker layer, samples were prepared by
pouring a mixture of 2:1 Tempera to water onto an adhesive functionalized piece of neoprene rubber.
The sample was held vertical and the excess mixture was allowed to drip off until the dripping had
reduced to less than 1 drop per second. To stretch the samples, they were loaded into a vice and slowly
pulled until cracks were visible under a microscope at 45 times magnification. After the pull test, a
failure was induced at the edge with a razor blade and the sample was torn with pliers. For each
sample, the coating tore readily and did not stretch across the tear.

Tempera
Sample # Thickness
355
0.001
356
0.002
357
0.002
358
0.001
359
0.001

Initial
Length
1.625
1.25
1.5
1.75
1.625

Length when
cracks formed
2
1.375
1.75
2
1.875

Strain
23%
10%
17%
14%
15%
Figure 73: 45x microscope image of
sample 357 in the strained state.

Every sample met or exceeded the 10% strain requirement. Furthermore, the sample appeared to be
well coated. Before straining, the rubber backing was not visible through the coating. However, when
the 2:1 tempera : water mixture was poured onto a sensor sample as a dielectric, it did not insulate well.
In fact, the resistance between the electrodes was less than or equal to 100k for every sample.
For the next test, samples were prepared by pouring a mixture of 3:1:2 Tempera : Latex : Water onto
adhesive functionalized neoprene coupons. The samples were then loaded into a vice and pulled until
80

cracks were visible under a microscope with a magnification of 45x.


Each of the samples strained to the full capacity of the vice (>100%
strain) and showed no signs of cracking. When a local failure was
induced with a razor blade and the material was torn with pliers the
film tore readily and did not stretch across the tear in the rubber.
In light of the promising results of the strain test, samples were

Figure 74: 45x microscope image of


sample 368, a 3:1:2 mixture of
tempera, latex, and water
respectively

prepared by pouring a 3:1:2 mixture of tempera : latex : water onto an adhesive functionalized
neoprene rubber sample with layers of RC200 electrode paint above and below. Each layer was
allowed to cure before the next layer was applied. The tempera mixture was used as the dielectric
isolating the electrodes. During the manufacture of these samples, it was noticed that the electrode
paint shriveled and wrinkled as it dried on the 3:1:2 mixture described above, but the reason for this
was not determined. Another interesting observation was that the dielectric layer separated from the
electrode and formed a thin, rubbery film.

SN
360
361
362
363
364

Final
Insulated? Failure Mode Resistance ()
Yes
Pinhole
30k
No
NA
NA
Yes
Tear Out of Plane
20,000k
Yes
Tear In Plane
8k
No
NA
NA

Figure 75: Electrode paint wrinkling where it


makes contact with 3:1:2 tempera : latex :
water paint

81

In-Plane Tear
Damage

Simulating
Burst Failure

Figure 76: Change in resistance between the two electrodes as a result of in-plane tear damage

6/20/2012
As a follow up test, sensors were prepared by pouring the electrode paint RC200, allowing it to dry,
then pouring on a dielectric paint. The sample was held vertical to allow excess to drip off until there
was more than one second between each drop. This method was used to remove excess from each of
the layers. The dielectric paint used was a mixture of 3:1:2 parts tempera : latex : water. The dielectric
layer was allowed to cure fully before the second coat of the same dielectric mixture was applied. The
second coat of dielectric was allowed to fully cure before the second layer of RC200 electrode paint was
applied and allowed to cure fully.
The samples were still not fully electrically insulated (>40M), but were significantly more consistent
than the single coat of the same dielectric paint. A measurement between the exposed portion of the
dielectric and the exposed portion of the bottom electrode indicated that the dielectric paint was not
fully insulating (which ruled out the possibility of the two electrodes making contact at the edges). Still,
a significant (at least one order of magnitude) decrease in resistance was measured as a result of each of
the three damage modes tested.

82

Initial
Final
Sensor # Resistance (k) Failure Mode Resistance (k) Notes
381.1
7,400
Pinhole
2.7 / 7.0
Steel Probe in / Steel probe out
382.1
5,800
Tear out of plane
480
383.1
3,000
Tear in Plane
125
384.1
9,000
Pinhole
2.0 / 6.0
Steel Probe in / Steel probe out
385.1
7,000
Tear in Plane
8,500

Figure 78: Tear in plane failure of sensor 383.1

Figure 77: Resistance between the electrodes following


the tear in plane damage

Following up on the results of sensor samples 332 through 347 (the batch of 16 sensors made with 4
coats of dielectric sprayed on) and the relative success of the pouring method, sensors were prepared
by pouring the electrode paint RC200, allowing it to dry, then pouring on a dielectric paint. The sample
was held vertical to allow excess to drip off until there was more than one second between each drop.
This method was used to remove excess from each of the layers. The dielectric paint used was a mixture
of 3:1 parts tempera : latex without water. The dielectric layer was allowed to fully cure before the
second layer of RC200 electrode paint was applied and allowed to cure fully.
The samples were still not fully eclectically insulated (>40M), and, in fact, created a worse insulator
than the 3:1:2 mixture previously tested. A measurement between the exposed portion of the dielectric
and the exposed portion of the bottom electrode indicated that the dielectric paint was not fully

83

insulating. Still, a significant (at least one order of magnitude) decrease in resistance was measured as a
result of each of the three damage modes tested.

Initial
Final
Sensor # Resistance (k) Failure Mode Resistance (k) Notes
386.1
1,500
Pinhole
1.3 / 3.3
Steel Probe in / Steel probe out
387.1
2,300
Tear in Plane
500
388.1
6,800
Tear out of Plane
8,800
389.1
4,300
Pinhole
1.1 / 4.1
Steel Probe in / Steel probe out
390.1
7,100
Tear in Plane
46

Undamaged

Probe In

Probe Out

Figure 79: 3:1 tempera : water dielectric poured onto the sensor without being thinned. Progression of pinhole damage
(sensor 389.1)

6/20/2012
After taking note that the rubber supplied by Parker Hannifin (Lot: 3155R1) was quite thin and relatively
soft, the idea came to use that rubber as a dielectric. Since the hoses are vulcanized and the
vulcanization process changes the chemical composition of the rubber, several samples were cut and
placed in the isotemp oven for 1 hour at 310 F. The samples were placed in the oven in different
configurations: flat, stretched, and hung. Flat indicates that the sample was simply laid onto foil in the
oven. Stretched indicates that the sample was stretched to a point just before its breaking point and
held there for approximately five seconds before being placed on the foil (where it proceeded to retract
to a state close to its original dimensions). Hung means that the samples were suspended from a wire
so that both sides were exposed to the oven air.

84

After vulcanization, the rubber was allowed to cool and RC200 electrode paint was poured onto one
side, taking care not to get any on the edge, and allowed to dry completely under a fume hood. The
drying process of the RC200 paint caused the samples to flex in a convex fashion. Once the first side was
dry, the samples were flipped over and RC200 was poured onto the other side with the same care to not
let the paint touch the edge and the sample was allowed to dry under the fume hood.

Hung

Hung

Flat

Flat
Flat
Stretch
ed

Flat
Stretch
ed

Figure 80: Parker Hannifin 3155R1 rubber samples in the oven for vulcanization

85

Stretch
ed

Final
Vulcanization
Sensor # Insulated? Fialure Mode Resitance (k) Configuration Notes
380
Yes
Pinhole
Insulated
Flat
381
Yes
Tear In Plane
2.3
Flat
382
Yes
Tear Out of Plane
1.4
Flat
383
Yes
Pinhole
.4 / insulated
Flat
Steel Probe In / Steel Probe Out
384
Yes
Tear In Plane
Insulated
Flat
385
Yes
Pinhole
.5 / Insulated
Stretched Steel Probe In / Steel Probe Out
386
Yes
Tear In Plane
Insulated
Stretched
387
Yes
Tear In Plane
Insulated
Stretched
388
Yes
Pinhole
.57 / 17,000
Hung
Steel Probe In / Steel Probe Out
389
Yes
Tear In Plane
Insulated
Hung
390
Yes
Foreign Object
300
Hung
Hit with edge of nylon cylinder

Figure 82: PH rubber dielectric after an


in-plane tear failure

Figure 81: PH rubber dielectric after an


out of plane tear failure

6/21/2012
In light of mediocre results from the sensors using Parkers lot 3155R1 rubber, additional tests were run
with this material to determine if better results could be obtained from a different batch. To prepare
these samples, Parker Hannifin rubber lot 3155R1. This rubber was chosen because it is thin and tears
easily. The blue protective backing was removed from the rubber and it was gently clamped to the table
with aluminum bars. RC200 paint was poured onto the rubber and spread thickly with a paintbrush.
Care was taken to ensure that the conductive material did not spill over the edge (though this was not
always successful). The RC200 was allowed to dry thoroughly (approximately 30 minutes) before the

86

samples were turned over and the process was repeated on the opposite side of the rubber. Any
sample that has conductive material spill over the edge was trimmed with a razor blade.
The results for the second batch were different from the first because the two electrodes were not
insulated from each other. However, the response of the sensor sample to failure was no better.

SN
421
422
423
424
425
426

Initial
k
6,000
7,000
800
11,000
500
2

Failure Mode
Pinhole
Tear
Pinhole
Tear
Pinhole
Tear

Final k

Notes
6.5 / 6,000 Vulcanized, Steel Probe In / Out
4,500
Vulcanized
2.5 / 7.5 Vulcanized, Steel Probe In / Out
14
1.5 / 18
Steel Probe In / Out
10.5

6/21/2012
In light of good results from the samples prepared with two coats of the 3 part tempera, 1 part latex, 2
parts water mixture, many additional samples were prepared to determine the repeatability of this
recipe. The sensors were prepared by pouring the electrode paint RC200, allowing it to dry, then
pouring on a dielectric paint. The sample was held vertical to allow excess to drip off until there was
more than one second between each drop. This method was used to remove excess from each of the
layers. The dielectric paint used was a mixture of 3:1:2 parts tempera : latex : water. The dielectric layer
was allowed to cure fully before the second coat of the same dielectric mixture was applied. The second
coat of dielectric was allowed to fully cure before the second layer of RC200 electrode paint was applied
and allowed to cure fully. Finally, three samples (SN: 403-405) were vulcanized at 310F for 1 hour to
inspect the sensors ability to withstand the vulcanization process.
All but one of the samples was insulated (>40M) and even the sample that was not fully insulated
showed a significant change (3 orders of magnitude) when damage was induced. Only one of the
samples tested did not show a significant change as a result of damage. The vulcanized samples
87

performed very similarly to the non-vulcanized samples.


SN
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410

Initial k
INS
INS

INS
INS
INS

INS
INS
INS
INS
INS
INS
INS
5,000

Failure Mode
Pinhole
Tear
Demonstration
Demonstration
Pinhole
Demonstration
Tear
Pinhole
Demonstration
Demonstration
Demonstration
Tear
Pinhole
Tear
Demonstration
Pinhole
Tear
Pinhole
Tear
Pinhole

Notes

6 / 8.3
70
2.5 / 5.0

80
5.5 / 29
340

Vulcanized
Vulcanized
Vulcanized

2 / 2.3
700
2.3 / 2.8
INS
2.6 / 4.7

Torn

Undamaged

Figure 83: 3:1:2 Tempera


Latex Water samples after
vulcanization

Final k
2.5 / 4.5
1,500

Figure 84: 3:1:2 Tempera Latex Water sample before and after tear damage

6/21/2012

88

The results of the sensor coupon samples that used 3 parts tempera, 1 part latex, and 2 parts water in
two coats consistently showed promising results. Therefore, it was decided to apply the sensors to
actual hydraulic hose samples to determine if these results carried over to a more realistic form. To
apply the sensors to hose samples, the ends were masked off with Teflon tape for a clean, crisp edge.
Teflon tape is used because it lacks adhesive which leaves a residue on the rubber and pulls the
dielectric from the electrode layer. RC200 was poured onto the hose and the excess was allowed to drip
off for ~5 seconds before the sample was rotated by hand steadily till the coat of RC200 has fully dried
(approximately 15 minutes). Teflon tape was used to mask off part of the RC200 layer on each side to
allow access to the electrode for testing. A mixture of 3 parts tempera, 1 part latex, and 2 parts water
was created and poured onto the sample. Again, the excess was allowed to drip off for about 5 seconds
before the samples were rotated continuously, by hand for a smooth, even coating. A heat gun was
used to expedite the process, but the drying time was still approximately 45 minutes. Once the first
coat of dielectric had cured, a second coat was added and the same drying procedure was used. After
the second coat had dried (again, approximately 45 minutes) another section at the end of each sample
was masked off with Teflon tape. A second RC200 adhesive coating was poured onto the sample,
allowed to drip for approximately 5 seconds and the sample was rotated continuously until the final
coating dried (again, approximately 15 minutes).
The images below show typical defects in the surface of the dielectric layer of the hose samples. The
bubbles were prevalent on all of the samples. However, the density of the bubbles varied from sample
to sample. There also seems to be a correlation between bubble density and initial resistance. For
example, there are relatively few bubbles in the image of sample 411 (which has the highest initial
resistance) and many more bubbles in the image of sample 414 which has a much lower initial
resistance. The bubbles may be related to the use of the heat gun to expedite the drying process.

89

Figure 85: Microscope image of


bubbles in dielectric on hose sample
411

Figure 87: Microscope image of


bubbles and tear in dielectric of hose
sample 412

Undamaged

Figure 86: Microscope image of


bubbles in dielectric of hose sample
414

Probe In

Probe Out

Figure 88: Pinhole failure test progression on hose sample 411

None of the samples were insulated (as defined by >40M). Also, only a pinhole failure was able to be
tested because the tough, steel reinforced rubber could not be torn by hand. However, every sample
showed at least a halving of the resistance from the pinhole failure. Most showed an order of
magnitude or greater decrease in resistance.
SN
411
412
413
414
415
416

Initial k
200
95
3
3
80
1.9

Failure Mode
Pinhole
Pinhole
Pinhole
Pinhole
Pinhole
Pinhole

Final k
.45 / .95
.45 / 1.4
.37 / .4
.6 / 1.5
.5 / 1.2
.4 / .6

Notes
steel probe in / out
steel probe in / out
steel probe in / out
steel probe in / out
steel probe in / out
steel probe in / out

3.4.5. Cost Analysis


The following is a rough cost analysis to provide a general idea of how much this sensor cost per inch of
application on a 1 inch diameter hose.
90

Item

Company

Dielectric
Liquid Latex
Tempera
Water
Electrode
Rubber Cement
Graphite Powder (635)
Toluene

3.4.5.1.

Amount Unit

Ben Nye
Aurora
NA

16
17
1000

Elmers
Southwestern
Whitaker

4.1
16
32

Cost

fl oz
fl oz
fl oz

$24
$2
$0

Amount / Cost/32oz
32oz mix
Mix
Reference (8/19/2012)
5.3
16
10.6

fl oz $2.29
oz
$12
fl oz $12

6.7
8
12.8
Total
Cost per inch lengh of 1" diameter hose:

$7.95
$1.88
$0.00

www.cosplaysupplies.com
www.misterart.com
NA

$3.74
$6.00
$4.80
$24.37
$0.25

www.staples.com
www.wayfair.com
www.allinonemarinesports.com

Electrostatic Deposition

6/25/2012
In an effort to achieve a thinner, more even coating of
dielectric material onto the electrode surface,
electrostatic deposition was considered. In
electrostatic deposition, a solution is loaded into a
syringe, brought to a voltage potential on the order of
kilovolts, and then steadily pumped (eg. at 40mL/hour)
through a very fine nozzle. When the highly charged
mixture escapes the confines of the nozzle, the atoms of
the charged liquid repel each other and create an

Figure 89: Schematic of electrostatic coating


machine. Image courtesy of the National
Engineering Research Center for Nanotechnology,
Shanghai, People's Republic of China accessed
June 28, 2012

atomized spray. The target of the deposition is grounded such that the spray is attracted to it. The
combination of the charged, atomized spray and grounded target ensures that the coating is attracted
most strongly to any uncoated portion of the target. This tendency for the spray to seek out uncoated
areas helps ensure that the coating is applied evenly and smoothly.
For this particular test, samples were prepared on adhesive functionalized neoprene rubber coupons.
RC200 electrode paint was poured onto the coupons and the excess was allowed to drip off till there
was less than one drop per second. The coat was allowed to dry fully. A mixture of 5% PVC by weight

91

and tetrahydrofuran (THF) was loaded into the 10mL syringe of the electrostatic deposition machine.
The mixture was loaded into the syringe press and fluid was pumped through the system until the nozzle
no longer bubbled. Once a smooth, steady drip was coming from the ceramic tip of the machine, the
speed of the mixture extrusion was reduced to 40 milliliters per hour and the voltage was adjusted till
the drips coming from the head became a very fine mist (which occurred at 12 kV). A portion of the
electrode was masked off with Teflon tape and a grounding alligator clip was clamped to the electrode
of the sample to ground the sample and thus attracted the atomized spray of PVC. For most samples,
the deposition lasted approximately 15 seconds and produced a wet sheen on the sample. After
deposition, the samples were placed into an oven at 80C for at least 5 minutes to evaporate the THC.

Figure 91: Sample being electrostatically coated with


PVC

Figure 90: Resistance test showing that the PVC


had insulated the surface of the sample before
the second layer of electrode was applied

Figure 93: Resistance between electrodes of


sample with PVC dielectric. Sample is poorly
electrically insulated

Figure 92: Sample with steel probe


inserted to connect the electrodes.
Resistance increased.

92

SN
433
434
435
436
437
438
439
440
441
442

Initial
Failure Mode
Final
325
Pinhole
572 / 750
800
Tear
8,000
300
Pinhole
650 / 350
400
Tear
5,500
400
Pinhole
500 / 500
340
Tear
900
390
Pinhole
600 / 390
340
Tear
1,500
PVC was not altered by vulcanization
350
Pinhole
450 / 470

Notes
Deposited for a minute
Deposited for 30 seconds
Deposited for 15 seconds
Deposited for 15 seconds
Deposited for 15 seconds, Stretched
Deposited for 15 seconds
Deposited for 15 seconds
Deposited for 15 seconds
Deposited for 15 seconds

6/29/2012
In light of the very poor electrical isolation that resulted from a 15 second coat of 5% PVC in THF at a
flow rate of 40mL/hr, a series of thicker coats were attempted. For this test, 16 1"x3" adhesive
functionalized neoprene coupons were made. RC200 electrode paint was poured onto the samples and
the excess was dripped off until the drip rate was less than 1 drip/sec. The electrode was allowed to try
completely before the electrostatic deposition began.
During the electrostatic deposition, a mixture of 20% PVC completely dissolved in THF was used as the
coating material. The material was applied at a rate of 40mL/hr and the voltage potential between the
charged material and grounded sample was 8,000 volts. The coating was applied in 15 second coats.
After each 15 second coat, a heat gun was used to evaporate any remaining THF from the samples
before the next coat was applied.
The initial insulation between the electrodes was very inconsistent and did not show a strong
correlation. When investigating the (k of insulation)/(seconds of coating), the standard deviation is
greater than the average even after the extreme outliers are removed. The inconsistencies of these
samples were also present in the final resistance measurements after damage. In some cases the

93

resistance decreased, in some cases it increased, and in some cases there was almost no change. These
samples were unpredictable.

SN
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466

Coat Time
(s)
15
15
15
15
30
30
30
30
45
45
45
45
60
60
60
60

Initial (k)

Failure Mode

Final (k)

300
7.5
34
82
52
11.5
30
105
32.4
10.7
40
18
32
Insulated
11
136

Pinhole
Tear
Pinhole
Tear
Pinhole
Tear
Pinhole
Tear
Pinhole
Tear
Pinhole
Tear
Pinhole
Tear
Pinhole
Tear

345 / 274
60
16.1 / 29
Insulated
68 / 95
126
10.7 / 22
168
12 / 405
1,800
20 / 13.6
197
17.6 / 31
270
14.1 / 49
825
MEAN
STANDARD DEV

k/(sec
coating)
OUTLIER
0.50
2.27
5.47
1.73
0.38
1.00
3.50
0.72
0.24
0.89
0.40
0.53
OUTLIER
0.18
2.27
1.43
1.46

Avg. per
section
2.74

1.65

0.56

0.99

When the samples were examined under a microscope (images not available because the reflective
graphite was visible through the thin PVC coating and severely interfered with the image quality), the
PVC coating was fibrous. There were very small gaps throughout the coating. Furthermore, it is likely
that the toluene in the conductive paint was reacting with the PVC.

3.4.5.2.

Composite Release Paper

6/28/2012
Most of the sensor skin designs had focused on a solid film dielectric. The disadvantage to this model is
that poor response to pressure (the simplest way to look at foreign object damage) is expected. Dr.

94

Schulz came up with the idea that separating the electrode layers with many small pillars would allow
the electrode paint to deform and make contact under pressure. In this way, the sensor could also
double as a type of pressure sensor. The ideal method of making this type of sensor is to apply the
dielectric with a mask that has been etched (for example, by photolithography) to allow extremely small,
cylindrical pillars to separate the electrodes. This method was not available, so porous release paper
was used as the dielectric instead. The release paper is intended for use in the manufacture of
composite materials to allow the cured composite to be removed from the mold. This paper is a
polymer film reinforced by woven fibers.
To construct the sensors, sections of the paper were cut and the RC200 electrode paint was poured onto
one side. The paint was allowed to dry completely before the sample was turned over and the paint
was poured onto the other side. The paint was, once again, allowed to dry fully (~5 minutes). The
standard pinhole and tear tests were run on the samples. In addition, a pressure test was run. During
the pressure test, the sensor was placed in a hydraulic jack between two pieces of 4"x4" wood. Some of
the senor hung out so that a measurement could be made while the sensor was under pressure. The
assembly was then loaded to 10,000 lbs. Since 4"x4"s are actually 3.75"x3.75", the area is
approximately 14in2 to distribute the force. Therefore, the sensor was under 711 PSI pressure. At this
pressure, the wood began to (audibly) crack internally.

Figure 94: Pressure test sequence used on the composite release paper samples

95

Unfortunately, none of the tests run produced any measureable response.


SN
443
444
445
446
447
448
449
450

Initial
Insulated
Insulated
Insulated
Insulated
Insulated
Insulated
Insulated
Insulated

Failure Mode
Pinhole
Tear
Pinhole
Pressure (711 PSI)
Tear
Pinhole
Tear
Pinhole

Final
Insulated
Insulated
Insulated
Insulated
Insulated
Insulated
Insulated
Insulated

Following these results, the composite release paper was


inspected further in an effort to determine the reason for
this lack of response. Under a microscope, it was clear that
the composite release paper, in fact, was not porous, but
simply textured. The polymer film coating the reinforcing
fibers also formed a film across the gaps between the

Figure 95: Microscope image of composite release


paper against a black background

fibers. The image below shows the composite release


paper against a piece of paper colored black with a permanent marker. Against the dark background,
the film between the fibers can be seen clearly.

3.5.Prototype Tests
After much development, prototype hoses were constructed and tested. The sensor skin was applied to
actual hydraulic hoses and put through pinhole, impact, bending, burst, and impulse tests. The results
of each of these tests are discussed below.

96

Figure 97: Applying electrode to hose


samples

Figure 98: Applying dielectric to hose


samples

Figure 96: Hose samples curing on


mandrel

The tests were run using:


NI 9219 - resistance measurement
NI 9163 - Wi-Fi single module carrier for 9219
NI 9921 - Ingress protection (IP) enclosure for 9219 & 9163, and its accessories
NI 9234 - vibration measurement
Kistler 8763B1K0AB Impact Accelerometer

Figure 99: Schematic of Data Acquisition

For more thorough detail regarding the configuration, operation, and verification of the data acquisition
system, please see Appendix 9.3

3.5.1. Response to Pinhole


Damage
The pinhole test was performed similarly to
other the previous pinhole damage tests. A
sharp steel probe was used to penetrate the

97

Figure 100: Typical resistance vs. Time Plot for Pinhole Damage

hose rubber through to the steel braid and then removed. The resistance between the electrodes was
recorded at a rate of 2Hz while the test was being run. The initial and final resistances were also verified
using a Fluke multimeter. In all 8 samples tested, the resistance immediately dropped to half or less
(sometimes several orders of magnitude less) as soon as the sensor was damaged. This significant drop
in resistance indicates that pinhole damage as occurred.

3.5.2. Response to Impact Testing


Impact testing was also performed using the
custom built impact test apparatus at the
University of Cincinnati. During this test, hose
samples were clamped onto the test apparatus and
an 8lb ax was allowed to swing from 30 inches of
elevation down onto the sample. An
accelerometer measured the acceleration of the ax
as it impacted the sample (the data was then converted

Figure 101: Typical impact force data and resistance


response of the sensor.

to force based on the mass of the ax) and the resistance between the electrodes of the sensor was
recorded almost simultaneously. Some data acquisition timing error caused the samples to not line up
chronologically.
In every test that showed visible damage from the ax the sensor responded very strongly and
permanently. This is a good result because significant damage to a hose from a foreign object does not
heal itself. The sensor indicates that the damage has occurred. However, when a protective coating
was applied to protect the sensor, the response was not as significant or permanent. The signal from
benign impacts and damaging impacts are drastically different and distinct.

98

Figure 103: Hose sample before impact ax test

Figure 102: Hose sample after impact ax test

3.5.3. Response to Bending Testing


To date, the bending tests that have been performed are
qualitative. With the sensor attached to the data acquisition device,
a researcher held the hose sample and slowly bent it while data was
recorded. The data showed a piezoresistive response to bending.
Figure 104: Hose being bent while
the resistance is measured

3.5.4. Response to Burst Testing


While at the Parker Hannifin Hose Product Testing facility in Wickliffe OH burst tests were run under the
supervision of Mr. Jim Henighan. This test utilized full hose samples with crimped on end fittings. One
end of the hose was plugged with a screw-in plug. The hose was filled with water and the other end was
attached to the burst machine and the protective covering is lowered over the hose to protect the
operators. The device then gradually increases pressure until the hose bursts open.
The hoses used in this test had a working
pressure of approximately 2,300 PSI.
They burst at approximately 11,000 PSI.
Each of the hose samples that burst in a
location where the sensor was applied

Figure 105: Parker Hannifin Burst Test Apparatus

99

gave a clear signal when failure had occurred.

Figure 107: Data from burst test machine

The above plots show two different trends

Figure 106: Burst hose following burst test

Figure 108: Pressure and resistance


Figure 109:
values
Pressure
for burst
and resistance
test
values for burst test
sample 20
sample 14

from the burst plots. The first plot shows a


steady (though nonlinear) increase in resistance as the pressure increases followed by a very sudden
drop in resistance once the hose burst (as indicated by the sudden pressure drop). The second plot
shows little response to pressure change, but a clear increase in resistance when the hose bursts. These
two plots are indicative of the variability in this test. Some samples showed an increase in resistance at
failure while others showed a decrease in resistance at failure. However, all samples showed a very
sudden change in resistance. By windowing the data and applying the Fischer criterion, this sudden
drop could be easily identified.
This correlation between pressure and resistance may seem opposite of what is expected. However,
that expectation is derived from a model of the hose that would suggest it is
a homogeneous material. In reality, the hose is constructed with a braid angle such that when it is
pressurized, the hose increases in diameter but decreases in length such that it squeezes together. This
is designed into the process to allow the hoses to be pressurized then removed from the mandrel. This

100

squeezing effect would cause the thickness of the dielectric to increase which would increase the
resistance between the electrodes.

3.5.5. Response to Impulse Testing


Finally, the most time intensive, but most realistic, test was the impulse test. For this test, three hoses
were loaded into the same manifold and pulsed through a pressure cycle which simulates actual field
use. This test, under the assistance and supervision of Mr. Gary Johnson, was run at a pressure cycle
rate of Hz and started with 2,000 PSI (just below the 2,300 PSI working pressure). 500 cycles were
completed at this pressure before the pressure was
increased by 500 PSI and another 500 cycles were
completed. The pressure was continually increased by 500
PSI after each 500 cycle run until each hose failed. Once a
hose failed, that hose was removed and that spot in the
manifold was plugged up. If the hose failed in the middle of
a run, that hoses manifold spot was plugged up and the
rest of the hoses were cycled at the same pressure until they

Figure 110: Impulse hose failure

experienced 500 cycles at that pressure before the pressure was increased. This process was continued
until all three hoses had failed.
In total, 6 hoses were tested. 2 of the hoses
were undamaged, 2 were kinked to damage
the steel braids, and 2 were damaged with
an etcher to assist a pinhole in forming.
However, all of the hoses failed via burst and
Initial Impulse Test Setup
101
Figure 111: Initial hose setup for impulse test

no pinhole was recorded. Pressure pulses were evident in the data. Furthermore, once the hose burst
through the sensor, the resistance of the sensor reduced by at least half of the original value. This is a
clear damage detection with a possibility for cycle counting via a piezoresistive effect.

Hose Failure

Impulse Machine Off

Impulse Machine On

Figure 112: Long term data from hose impulse test showing machine shut-off and hose failure

102

Impulse Cycles

Hose Failure

Figure 113: Short term data from hose impulse test showing pressure cycles and hose failure

3.5.6. Response to Thermal Testing


The sensor was put through a simple thermal
test to investigate how the resistance of the
sensor would respond to significant changes in
temperature. To do this, an oven was brought
to 310 degrees Fahrenheit (155 centigrade) and
the sensor samples were placed in the oven.
The resistance of the sensor was automatically

Figure 114: Sensor samples in the oven with wires attached for
resistance measurements

recorded using a data acquisition device at a


sampling rate of 2Hz. The temperature of the sensor was recorded using an infrared laser thermometer
every 5 minutes. The temperature profile of the sensors is shown below.

103

Temperature profile of Sensor Samples


300

Temperature (F)

250
200
Sensor 1
150

Sensor 2
Sensor 3

100

Sensor 4

50
0
1190

1200

1210

1220

1230

1240

Time (military)
Figure 115: Temperature profile of the sensor samples

The resistance profiles for the sensors are inconsistent, as shown below.

Pinhole Failure

Figure 117: Full resistance data for thermal tests

Figure 116: Focus on induced failure data from thermal tests

104

The data from the first sample shows a very smooth progression with a clear signal when a pinhole
failure is induced. However, the data for the other samples is relatively random and not much
conclusion can be drawn from it. Further tests should be run to extract a meaningful conclusion about
the response of the sensor to temperature.

3.6.Problems and Concerns


The sensor skin concept has many advantages as discussed above. However, as with any technology,
there are problems that need to be addressed. One problem is prohibitive material costs. While costs
will vary with application, specifications, and lengths, a typical hydraulic hose is fairly inexpensive. An
inquiry to the Tractor Supply Company, a distribution company for farm equipment, indicates that a
diameter, 84 length hose rated for 3,500 PSI is only $24.99 (11). Hydraulic hoses are often farm
machinery, trucks, buses, and virtually all types of earth moving equipment (12). Even on the low end,
these machines are very costly starting at about $24,000 (for a Toyota Model 8FGCU25, 5,000 lb. lifting
capacity, powered by LPG, 3-stage mast with 189 of lifting height and a side shifting attachment with
42 forks (13). Therefore, while the hose itself is not a costly component, it is critical in maintaining the
health of operators and the very costly equipment it is attached to.
3.6.1. Intrinsically Safe Device
As a result of a discussion with Mr. Aaron Johnson, it was brought to my attention that in the oil and gas
industry equipment must be deemed intrinsically safe. Intrinsically safe equipment is equipment and
wiring which is incapable of releasing sufficient electrical or thermal energy under normal or abnormal
conditions to cause ignition of a specific hazardous atmospheric mixture in its most easily ignited
concentration. (14) There are three elements necessary for ignition and combustion: fuel, oxygen, and
ignition. Due to the nature of the oil and gas industry, both fuel and oxygen are inherently present.
Therefore, for a device to be intrinsically safe, it cannot be capable of building up enough charge to
produce a spark or generate enough thermal energy to reach the ignition temperature.
105

Both thermal and electrical energy limitations present challenges that may not be initially obvious. For
example, the thermal limit applies to any element of the device. Therefore, if the device has a very fine
wire with current running through it, the limit on the current must be very small because it will not take
much power to greatly increase the temperature of a fine wire. Additionally, electrical components
which are close to each other may pose a potential danger. Assume that the fine wire has very little
current going through it, but, after years of vibration, the wire fatigues to the point of failure and makes
contact with a nearby component that has more current. This new connection may bring the fine wire
to the ignition temperature.
Similarly, there are many hidden issues with sparks. For a spark to occur, a significant change must be
built up. However, even with a very small power supply, significant charge can accumulate if the
apparatus has capacitance.
The most significant threat to intrinsic safety inherent in the sensor skin design is the potential for
significant capacitance. The sensor skin is, theoretically, a very large parallel cylinder capacitor.
Assuming a strong dielectric, charge can build from fluid moving through the inside of the hose or water
moving across the outside. When this charge is great enough, a spark can occur and ignite any nearby
flammable material.
One solution to this problem is the utilization of a zener diode. A
zener diode behaves like an open circuit until the voltage potential
across the diode exceeds its rated breakdown voltage. Similar to a
pressure check valve in a fluid system, once the breakdown voltage
is exceeded, the zener diode allows current to pass through. Zener
diodes are readily available at electronic component stores and

Figure 118: Zener diode circuit courtesy


of Wikimedia Commons. Accessed
7/17/2012

have many options for breakdown voltages. While they are somewhat temperature dependent, a
106

breakdown voltage of 5.6V is an equilibrium point which allows the diode to be temperature
independent. By bridging the two electrodes with multiple zener diodes (utilizing redundancy for
safety), the possibly of significant charge buildup is eliminated.

3.6.2. Temperature Induced Embrittlement


The ductility of polymers is highly dependent on temperature. Each polymer has a temperature known
as the glass transition temperature at which point the polymer is no longer ductile but instead becomes
brittle like glass. When designing the sensor, it is important that the material of the sensor has a glass
transition temperature less than or equal to the glass transition temperature Tg of the material it is
sensing or lower than the temperature of the application. If the sensor has an unacceptably high Tg then
it will fail prematurely under normal
operating strain.
When determining the Tg of a
material, it is important to test the
material in an axial strain test. At
first it may seem appropriate to
simply bend a sensor sample. While
the sensor is strained as a result of
bending, the thin sensor on the
outside of the relatively thick hose

Figure 119: Determination of Tg by dilatometry. The linear sections below and


above Tg are marked green; Tg is the temperature at the point of intersection
of the corresponding red regression lines. Image courtesy of Afluegel

effectively experiences purely axial

107

strain at each infinitesimally small spot on the surface of the hose. However, the flexibility of a material
(such as the film of the sensor) with respect to bending is a function of the product of both the elastic
modulus and second moment of area.
(15)
Where M is the bending moment, R is the radius of curvature, E is the elastic modulus, and I is the
second moment of area. For this application, it is important to focus on R and I. If one mistakenly
assumes that a small radius of curvature on a thin sensor skin sample will relate to a small radius of
curvature that the sensor can survive on a hose, then there will appear to be a mismatch in the data.
The reason for this is the difference in I.
For the sensor skin:

Where b is the width of the sensor, but, more importantly, h is the thickness of the sensor. If the sensor
is very thin, this number will become very small allowing for a very small radius of curvature. However,
as noted earlier, the primary strain that the sensor film will experience is an axial strain.
As the elastic modulus changes with temperature, the strain to failure will also change. These changes
must be kept in mind for extreme temperature applications and appropriately tested.
3.6.3. False Trigger Due to Abrasion
During the course of normal operation, a piece of equipment will experience rubbing and abrasion. A
simple example of this is a hydraulic power hose on a truck which swings and slides against itself and the
body of the truck as the truck drives down the road. Excessive abrasion is cause for a genuine trigger,
but a moderate amount of abrasion does not significantly affect the health of the hose.

108

The threat of an abrasion-induced false trigger is most significant on a surface sensor. The sensor can
experience abrasion before the material it is sensing has any damage. A sensor that is embedded in the
material or applied as a layer between hose windings would be less prone to this damage. One of the
primary goals of condition monitoring is to maximize the useful life of a structure. A premature sensor
trigger inhibits this effort.
3.6.4. Strain to Failure
All materials experience strain and deformation during any change in force or temperature. However,
not all materials strain in the same way. To accentuate this point, we will use two common materials as
contrasting examples: steel and latex. Steel is a relatively stiff material requiring a large force for small
deformation of steel while latex is not stiff and a small force will result in a very large deformation. If
these materials are bonded together in layers, for example steel latex then steel, and only one of the
steel layers is strained, the latex layer is soft enough that it will strain as well but not impart a significant
strain in the second steel layer. See the figure below for a graphical explanation of this.

Tensile force applied to lowest layer


Figure 120: Strain induced on layers of materials with very different stiffnesses

Another problem with material property mismatch is that one material may fail well before another.
Continuing with the steel and latex example, latex can strain over 700% before failure. Steel will fail well
before that point. Therefore, in the event of failure in the material, the steel will fail and break and the
latex will continue to stretch across the break. In the application of a hose sensor, this can create a
problem if the sensor fails at a different point than the hose. For example, in a hose rated for 10%
strain, it is important that the sensor fail at approximately 10%. A sensor that fails too quickly will

109

produce false alarms of failure during normal hose operation, but a sensor that can stretch too far will
not alert an operator when the hose has been compromised by excessive strain.

3.7.Potential Solutions to False Positive


A false positive occurs when a sensor indicates a failure when, in fact, none has occurred. This may be
due to material fatigue, a glitch in the software, a short in a wire, etc. but regardless of the cause, a false
positive can result in unnecessary product waste and increased cost in products and maintenance time.
One possible solution to false alarms by both abrasion and glass transition temperature is to use
redundancy. Using multiple layers of electrode and dielectric is one way to implement redundancy.
Multiple, independent layers allows for redundant contact points between the electrodes to sense
failure as well as redundant wiring to ensure that the sensor stays well connected to the monitoring
chip.
Another approach to redundancy is to use several independent sensors at different layers on the object
being monitored. For example, one sensor along the inner wall of the hose and another sensor
encasing the outside of the hose. For fluid to exit the hose, both sensors would need to fail. However, a
connection between the electrodes of only the inside sensor would indicate the propagation of a
pinhole leak. A connection of only the outside sensor may indicate degradation due to abrasion or
strain. Both abrasion and strain are accentuated on the exterior of the hose. A failure of both sensor
layers simultaneously would likely indicate a kink or burst.
Redundancy can also help resist false positives that result from dropping below the glass transition
temperature if the sensor is placed at intermittent layers of the hose. For example, a hose that is
carrying liquid nitrogen gas (LNG) in an ocean will have a very low temperature on the inner core where
the LNG is flowing through, but will be a more moderate temperature on the outside of the hose
because the temperature of the ocean cannot drop far below the freezing point of water. Therefore, a
110

sensor on the outside of the inner core would be much colder than a sensor just inside the outer
carcass. Still, before any fluid could escape the hose it would have to pass through both sensors.

111

4. Home Made Devices


4.1. Impact Tester
One of the common failure modes for
hydraulic hoses (especially HPD hoses) is
foreign object damage. To date, the only
measures taken to address this problem
come in the form of coatings and sleeves
intended to protect the hose such as
Parkers TOUGH COVER and SUPER TOUGH
COVER (16). However, even with these
covers, the structural integrity of the hose
can be jeopardized without any clear signs
to the operator.
Following a visit the Parkers test facility in Wickliffe,

Figure 121: Parker Hannifin's impact test apparatus

Ohio a fairly simple test apparatus was reviewed and an apparatus similar in function was designed and
built at the University of Cincinnati. The Parker apparatus consists of a rectangular support structure
with two cylindrical steel shafts for linear bearings to ride along. Mounted to the linear bearings is a
structure which holds weight (up to 15 lbs) and can incorporate interchangeable tips. Two knobs (one
on each side) can be pulled simultaneously to cause the weights to drop down (approximately 15 inches)
onto a hose sample. The hose sample is clamped to the test bed on two sides of the impact location.
At UC, a similar test setup was made which incorporates the same basic functionality of Parkers test
apparatus with some design insight taken from the Charpy test. The UC impact tester uses a 3ft, 8lb
maul mounted to a steel shaft which rides on ball bearings. An impact accelerometer is mounted to the

112

back of the hammers head to measure the acceleration of the head as it strikes the sample. This
measurement provides a quantitative impulse profile of the strike. With a full swing, the UC impact test
apparatus can deliver up to 40.0 joules of energy to the sample while Parkers impact test setup can
impart approximately 13.5 joules of energy into the hose.

Figure 122: University of Cincinnati's Impact Tester

Item
1/2" Diam. Mounted Bearing
1/2" by 12" Drive Shaft
1" Single Angle Bracket
Fasteners
Aluminum T-Slot Framing
8lb Splitting Maul

Company
McMaster
McMaster
McMaster
McMaster
McMaster
Menards

Part Number
6244K51
1346K17
47065T223
47065T139
47065T101
TJ-8F-MEN

Qty
2
1
18
10
4
1

Price
$36.77

$7.22
$3.98
$1.85
$14.20
$27.47

Total

113

Sub Total
$73.54
$7.22
$71.64
$18.50
$56.80
$27.47

$255.17

The transducer, cabling, and DAQ are not included in the parts list because they were available in
laboratory stock through Dr. Jay Lees Intelligent Maintenance Systems Laboratory.

4.2. Mandrel
After the first batch of sensors were painted onto hose samples it was noticed that the application
process was very time consuming as a result of drying time. Furthermore, while the sensor was drying
the sample needed to be continuously rotated to ensure that the sample dried evenly. Therefore, each
sample required almost 2 man hours.
To run a reasonable number of tests, a simple device was made that the hoses could mount onto and be
steadily rotated at approximately 100RPM without the constant supervision of a researcher. The device
is powered by a geared AC motor connected to a 5ft shaft by a rigid coupling and connected to two
other shafts by timing belts and cogs. The ends of the shafts are seated in ball bearings to assist with
rotation and reduce wear.

Figure 123: Mandrel constructed at UC

114

Item
12"x24" steel sheet
3/8" bore flange-mount bearings
3/8" bore 10T timing belt pully
1/2" wide, L size timing belt
Mechanical Coupling 8mm Bore
Femail Crimp on Quick Disconnect
Male Crimp on Quick Disconnect

Company
McMaster
McMaster
McMaster
McMaster
McMaster
McMaster
McMaster

Part Number
6544K14
4575N35
6495K23
6484K144
2469K2
7243K11
7060K25

Qty
2
9
6
3
2
2
1

Price
$16.70

$15.20
$16.65
$7.66
$28.68
$11.41
$11.36

Total

Sub Total
$33.40
$136.80
$99.90
$22.98
$57.36
$22.82
$11.36

$384.62

The motor is not included in the parts list because it was available in Dr. Schulz Nanoworld lab.

115

5. Electroactive Polymer Sensor


An electroactive polymer (EAP) sensor is thin, insulating polymeric sheet with a conductive coating on
each side. For example, a silicone-based polymer with a silver coating on each side. The silver coating is
very thin to ensure that it can flex and strain with the polymer but is still conductive.

5.1.Electroactive Polymer Theory

Figure 124: Electroactive polymer sensor courtesy of


http://lmts.epfl.ch/files/content/sites/lmts/files/shared/images/DEAP_principle.jpg Accessed 4/9/2012

The above figure illustrates the active element of an electroactive polymer. When a high voltage is
applied, the electrostatic force between the electrodes will pull the electrodes together squeezing the
insulating polymer between the electrodes. As a result of Hook's law, the insulting polymer, and thus
the thin silver coating, stretch out to the sides increasing the surface area of the electrodes.
A similar phenomenon can be used in conjunction with capacitance measurements to detect strain.
When the EAP sensor is strained in the plane of the senor the surface area of the electrodes increase.
Two conductive surfaces separated by a dielectric form a capacitor. An increase in the surface area and
decrease in the thickness of the dielectric increases the capacitance of the material. Therefore, an
increase in the capacitance of the sensor indicates an increase in the strain.
116

5.2.Electroactive Polymer Models


EAP sensors can detect strain in the plane of the sensor. However, it is incapable of discerning the
specific direction of the strain. For example, a square sensor that is placed in the x-y plane can detect
strain in the x-y plane and not the z direction, but the sensor carries no information to determine if the
strain is in the x direction or the y direction. This problem has been encountered before in strain
measurements by metallic foil strain gauge developers. Metallic foil strain gauge developers overcame
this issue by directing the sensor overwhelmingly in one direction by means of a snaking line that follows
one direction much more prominently than another direction. This sensor design allows the sensor to
be two orders of magnitude more sensitive to strain longitudinally along the gauge than perpendicular
to it. The effects of both strains are still present, but significantly reduced.
The same solution can be applied to EAP sensors by reshaping the sensor into long strips. In a hose
application there are effectively only two directions of significant interest: longitudinal direction along
the length of the hose and the circumferential direction along the circumferential cross section of the
hose. When a hose is pressurized, it will strain in both directions, however, in bending, the change in
strain is only longitudinal. The inside of the bend will compress while the outside of the bend will
stretch to allow for the curvature. Monitoring the circumferential strain is a good means of determining
pressure strain in the hose and is also necessary for factoring out the pressure strain to accurately
detect longitudinal strain that would result from bending.
Alternatively, it is possible to detect both pressure and bending strain from longitudinal sensors alone.
To explain this, consider a simple example of a hose that is bent around some radius of curvature. In the
plane of the bending there is an EAP strain sensor on each side of the hose (along the inner diameter
and outer diameter). If we allow C to be the base capacitance of the sensor, the capacitance after the
hose is pressurized, Cpressure, out, will be larger than the original C on the outside of the bend because the

117

sensor has strained causing the surface area of the electrode to increase and the thickness of the
dielectric to decrease. Similarly, capacitance of the sensor on the inside of the hose, Cpressure, in, will
decrease. However, if the hose is pressurized there will be a longitudinal elongation as well that will
increase the capacitance of the sensor.
These capacitances, Cpressure, in and Cpressure, out, form the upper and lower limits of the sensor capacitance.
The pressure increase will increase both the inner and outer capacitances equally and thus shift both
readings toward an increased capacitance. From this initial shift, the bending will cause the inner and
outer sensor capacitances to deviate from the new baseline. By comparing the inner and outer
capacitances the strain from bending and pressurization can be determined. The figure below helps
illustrate this point.
Inner Sensor
Outer Sensor
Strained Baseline
Capacitance
(C)

Baseline

Unstrained

Bending

Pressurized

Pressurized with
bending

Figure 125: Superposition of capacitance in EAP sensors

Notice that the strain from pressure and bending each influence the capacitance, but that the influence
of each can be separated by analyzing how far each sensor deviates from its baseline. A pressure strain
should influence both of the sensors similarly while a bending strain will have a different effect on each
sensor.

118

5.3.Experimental Data
Experiments were performed to determine the sensor response of the EAP material as a strain sensor
and puncture sensor. The first experiment was set up to determine the capacitance response of the
sensor to strain. This was done on December 15, 2011 using a Mark-10 tensile test machine (model
ESM301, SN 2188513) to accurately and smoothly strain the material and a Solartron Impedance Phase
Gain Analyzer (model SI1260). At incremental strains, the material was put through a frequency sweep
of 10Hz to 10MHz by the Solartron and the response was measured. One alligator clip was attached to
each side of the material. To protect the material from the teeth of alligator clip teeth and to ensure
connection and insulation and thin piece of copper sheet and folded piece of paper were used
respectively. Capacitance was calculated using the following equation:

The following and table show the results of this experiment.

Strain
1.000
1.028
1.056
1.084
1.112
1.140
1.168
1.196
1.224
1.252

Solartron
Capactiance
(nF)
6.5
6.6
6.7
6.8
7.0
7.1
7.3
7.4
7.6
7.6

Handheld
Capacitance
(nF)
9.5
9.5
9.0
8.7
7.3
6.3
5.0
3.5
3.0
12.0*

*Final measurement taken after material sat strained for three days. Indicative of creep effects, not strain effects

The mismatch in the capacitance value and, more importantly, in the trend of the capacitance of the
material is a point of concern. The Solartron data follows the trend that was expected based on the
physics model while the handheld data follows an inverse trend.
119

1.00E+07

EAP Sheet Amplitude Plot at Various Strains

1.00E+06
1.000
1.00E+05

1.028

Amplitude

1.056
1.00E+04

1.084
1.112

1.00E+03

1.140
1.168

1.00E+02

1.196
1.00E+01

1.224
1.252

1.00E+00
1.00E+00 1.00E+01 1.00E+02 1.00E+03 1.00E+04 1.00E+05 1.00E+06 1.00E+07
Frequency (Hz)

1.00E+02

EAP Sheet Phase Plot at Various Strains

8.00E+01
6.00E+01

1.000
1.028

4.00E+01
Amplitude

1.056
2.00E+01

1.084

0.00E+00
1.00E+00 1.00E+01 1.00E+02 1.00E+03 1.00E+04 1.00E+05 1.00E+06 1.00E+07
-2.00E+01

1.112
1.140
1.168

-4.00E+01

1.196

-6.00E+01

1.224
1.252

-8.00E+01
-1.00E+02

Frequency (Hz)

120

6. Fiber Optic Sensor for Strain and Temperature Monitoring


In ideal applications, fiber optic sensors provide a nonintrusive, highly sensitive means of measuring
temperature or strain at many points along an object. Furthermore, fiber optic sensors are insensitive
to electromagnetic interference, do not create a source for spark to occur, and are lightweight.
However, at approximately $150 per sensor, they can be cost prohibitive. (1)

6.1.Fiber Optic Theory


To measure strain and temperature, fiber optic systems typically use fiber Bragg gratings (FBG). These
gratings are photo-etched into the fiber and create a local change in refractive index. By spacing these
grating periodically and within the wavelength of light they can act as a local bandstop optical filter. (1)
When the spacing between the grating changes due to strain or temperature, the wavelength that is
filtered changes. By measuring the reflection spectra, the change in length can be determined and thus
the strain.
To use a single fiber optic cable for several temperature or strain measurements one can simply focus
on a different section of the light spectrum at different locations along the fiber. For example, imagine
the goal is to take 3 temperature measurements that are 10m apart. The FBG at the beginning might
filter out the red band of light, the middle filters out the green band, and the end filters out the violet
band. By reading the changes in EXACTLY which wavelength of each section (red, green, violet) is
filtered out, it is possible to determine what the temperature is at each section of the fiber.
The fiber optic system can consume over 6 watts of power through thermoelectric temperature control
of sensitive devices and powering the SLED (1). This power consumption is not a serious concern if a
power source is readily available, however if the system is in a place that is difficult to supply power to it
may require batteries or energy harvesting for power. In either of these cases a power budget becomes
critical.
121

Figure 126: Fiber Bragg grating multiplexing. Image courtesy of Energy Harvesting for Structural Health Monitoring Sensor
Networks, Los Alamos National Lab, 2007

6.2.Implementation into a Hose


Fiber optic sensors are primarily beneficial for two types of measurements: temperature and strain.
Both of these can be used when detecting leaks or damage in a hose. For large scale damage such as a
burst, blowout, or significant foreign object damage, there will likely be a change in stress
concentrations which will lead to a change in strain in the material. A single fiber optic sensor can be
wrapped around the hose in a helix to measure the circumferential strain. Circumferential strain can
indicate pressure in the hose and local damage. Additionally, fiber optic sensors can be run along the
length of the hose. Axial sensors are ideal for measuring bending strain in the hose as well as any local
damage. Overall, strain measurements are very useful for measuring structural damage to the hose.
However, strain measurements cannot detect relatively small leaks. For these, a temperature sensor is
more appropriate. Typically, the contents of a hose (e.g. hot hydraulic oil or cold liquid nitrogen gas)
produce a significant temperature gradient through the hose. In the event of a leak, the local

122

temperature will change significantly. This local temperature change can be detected by the fiber optic
sensor to communicate that a leak has occurred.

6.3.Problems and Concerns for Fiber Optic Sensors


Fiber optic sensors are good for measuring both temperature and strain; however, they are unable to
independently decouple the two measurements. For this reason, additional sensors must be used. One
option is to incorporate a dummy sensor which is not bonded to the hose, but run alongside another
sensor. This dummy sensor, because it is not bonded, will not be affected by strain but will still detect
temperature. By comparing the dummy sensor to the bonded sensor in the same vicinity, it is possible
to determine what portion of a measurement is from temperature and what is from strain. Similarly, a
thermocouple can be used to exclusively measure the temperature because thermocouples have a
negligible response to strain.
Another issue when implementing a fiber optic sensor into a rubber hose is stiffness mismatches. Silica
is a common material for fiber optic lines (17) with a stiffness of 57GPa (18).
Another issue with fiber optic sensors is the cost. The acceptable cost level for a sensor on a product is a
function of both the product cost and the value of information that the sensor is providing. For
example, a sensor that will monitor a hydraulic hose on a fork lift is not actually protecting the hose.
The sensor is, in fact, protecting the fork lift, operator, and whatever the lift is lifting. So, while the value
of the hose is likely less than $100, the potential damage that could result from that hose failing is much
greater. Still, if the hose is less than $100, it is a disposable part and can be replaced regularly for a
relatively low cost. However, individual fiber optic sensors are $100 to $250 each and the interrogators
can range from $12,000 to $45,000 according to Joe Thomas from Micron Optics on November 22, 2011.
Such a high price tag on a previously disposable component would be prohibitively difficult to sell to a
customer.
123

7. Conclusions and Future Work


Several different sensors have been explored in this study. Each sensor type was examined for several
attributes and compared against the other sensors. The following tables graphically display the
conclusions of this comparison.

Sensitivity to Root Cause

ter
De
le
ho
Pin

rai

o
ati

t St

ior

n
die
Gra

ing
O-r

ure
n
sio
rro
Co

ing

e)
ittl

)
ave

Ag

(Br

e (W

ure

igu
Fiber-optic
EAP
Thermocouple
Strain gauge
Pressure Sensor
Piezoelectric (clip-on)
Sensor Skin

rat
pe
Tem

rat
pe
Tem

Fat

Sensors

Medium

Good

Bad

Bad

Good

Bad

Medium

Medium

Bad

Bad

Bad

Good

Medium

Good

Bad

Good

Bad

Bad

Bad

Bad

Medium

Good

Bad

Medium

Medium

Good

Medium

Good

Bad

Bad

Medium

Bad

Medium

Bad

Medium

Good

Medium

Medium

Medium

Good

Medium

Good

Medium

Bad

Medium

Medium

Medium

Medium

Good

124

Attribute
rat
pe
Tem
siti
Sen
y
vit

y
ilit

n
tio

ise
No

a
ret

c
Pri

Fiber-optic
EAP
Thermocouple
Strain gauge
Pressure Sensor
Piezoelectric (clip-on)
Sensor Skin

rab
Du

p
ter

ure

In
nal

Sig

Sensors

Bad

Good

Good

Good

Medium

Good

Medium

Medium

Medium

Medium

Good

Good

Medium

Medium

Good

Good

Good

Bad

Medium

Medium

Medium

Good

Good

Medium

Good

Good

Medium

Medium

Medium

Bad

Good

Good

Good

Good

Good

The primary focus of this thesis has been on the sensor skin. The most significant conclusion is that it is
difficult for a mechanical engineer to refine material properties. Nonetheless, the final iteration of the
sensor skin with RC200 conductive paint and a dielectric that is 3 parts tempera, 1 part latex, and 2 parts
water performs well. This final iteration can regularly detect pinhole and tear damage by demonstrating
a significant change in resistance between the electrodes. The following table shows a comparison of all
the dielectric and electrode materials investigated during this study.

125

Material Concepts
Table
Spray on Rubber
Pure Latex
Latex/Thickener
Latex/Body Powder
Tempera
3 Latex: 1 Tempera
Scotchbright Sponge
Gel Medium
Latex Paint
Acryolite Paint
Liquid HNBR (Nitrile)
Silicone Rubber
Aluminum Foil
Latex/Graphite
Liquid Tape/Graphite
Rubber Cement/ Graphite
Medical Electrode

Electrical
Isolation
Medium
Good
Good
Good
Medium
Good
Good
Good
Medium
Medium
Good
Good
NA
NA
NA
NA
NA

Conductivity
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA

Flexibility
Good
Good
Good
Good
Medium
Good
Good
Good
Medium
Medium
Good
Good

Durability
Good
Good
Good
Good
Medium
Good
Good
Good
Medium
Medium
Good
Good

Chemical
Resistance
Bad
Good
Good
Good
Good
Good
Good
Good
Good
Good
Good
Good

Temperature
Sensitivity
Medium
Medium
Medium
Medium
Good
Good
Good
Medium
Good
Good
Good
Good

Proper
Failure
Good
Bad
Bad
Bad
Good
Good
Bad
Good
Good
Good
Bad
Bad

Comments
X - disolved w/ electrode
X - too elastic to fail
X - too elastic to fail
X - too elastic to fail
X - not elastic enough
Y
X - results mostly bad
X - not insulating
X - not insulating
X - not insulating
...investigating
...investigating

Good
Medium
Medium
Medium
Medium

Bad
Good
Good
Good
Good

Bad
Medium
Good
Good
Good

Good
Good
Medium
Medium
Medium

Good
Medium
Medium
Good
Good

Medium
Good
Medium
Good
Good

X - not flexible enough


X - latex clumped w/ graphite
X - rubber cement better
Y
may have other applications

The change in resistance can be detected by a computer using a learning algorithm. Such an algorithm
saves values over time and compares them to current values. A significant, especially sudden, change in
the resistance between the electrodes would indicate damage. Future work must be done to write this
algorithm and refine the thresholds to maximize responsiveness and minimize false positives.

7.1.Functionality
7.1.1. Proven
The sensor skin has been proven to be able to detect surface cracks and surface damage. Surface
damage severe enough to damage the sensor can also be detected. The sensitivity of the sensor can
easily be tuned by adjusting the materials used for each element of the sensor. Finally, the sensor has
been shown to respond to stress and strain via a piezoresistive effect.
7.1.2. Potential
There is great potential in the sensor skin. It is a very simple, rugged, versatile sensor with a huge
variety of applications. For one, the sensor most likely will have some thermal response. This may allow
it to be used as a thermal sensor. More importantly, however, is the potential for predictive monitoring.
The sensor has shown piezoresistive effects and has been tested for inductive and capacitive elements.

126

This full spectrum of attributes can be monitored during the life of a hose and, when combined with a
carefully written intelligent algorithm, may be able to predict when a hose will fail.

7.2.Thermoresistive Tests
Much of the potential functionality of the sensor depends on a piezoresistive effect. However, while
measuring the changes in resistance, it is important to fully understand the breadth of possible events
which may affect the resistance of the sensor. One very likely event that has not been tested as of the
writing of this document is the effect of temperature on the resistance of the sensor.
Hydraulic hose, drive belts, and tires all experience significant temperature swings between ambient
temperatures and working temperatures. Significant temperature swings will most likely have some
affect on the resistance of the sensor. It is important to determine how much affect temperature has on
the resistance of the sensor and how to work with or around these effects to create a usable product.
The ideal tests would involve a test of all of the failure modes at several temperatures along the working
range of the product. Furthermore, tests while the product is put through active temperature cycles
would be beneficial.

7.3.Additional Piezoresistive Tests


The sensor has demonstrated some piezoresistive capability. To maximize the usefulness of the sensor,
careful and extensive tests should be run to correlate motion of the substrate to response of the sensor.
The sensor has been shown to respond to pressure and bending in some way, but the time has not been
spent to relate the motion to the sensor reading. For precise measurements to be made, improved
process control during sensor manufacture may be necessary.

127

7.4.Life Testing
Most elastomeric devices have a high life cycle count. For this reason, the sensor must also be able to
survive a high cycle count without prohibitive degradation in performance. Though the sensor materials
were selected with longevity in mind, to date, no sensor has been cycled more than 10 times.
Both bending and pressure cycle tests must be conducted. The bending test can be performed at UC
using the pneumatic bending machine housed in the IMS lab. The pressure test, on the other hand,
must be completed at a professional testing facility for safety reasons. During both of these tests, data
should be taken continuously to monitor any sign of degradation which may develop before failure.

7.5.Additional Applications
At the core of its functionality, the sensor skin is a continuous, surface monitoring sensor. The ability to
continuously monitor large surfaces has application in many other fields as well. For example,
monitoring the entire surface of a concrete bridge could alert transportation engineers of possible
cracks and failures propagating from the surface of the support structure.
Similarly, the aerospace industry would benefit from continuous surface monitoring. Spacecraft are
continuously barded by micrometeorites that may puncture the vessel allowing oxygen to escape.
Several discrete panels of sensor skin would enable astronauts to narrow down where the damage had
occurred and repair it quickly. Airplane wings also have as very critical surface as they form the airfoil to
generate lift. Damage to this surface can cause significant decreases in fuel efficiency or invisible
internal damage to a composite wing.
For any application where the surface is critical, the sensor skin concept may find use. Specifically, the
paint-on application technique makes the sensor skin remarkably easy to add to almost any application.

128

The materials used for the electrodes and dielectric must be tuned to fit each individual application, but
the concept is valid.

7.6.Energy Harvesting
For some applications the sensor may be in a remote location or in a position where it is difficult to
deliver power. Many applications would benefit from long term operation that does not require
batteries to be replaced. Some remote applications may require wireless communication as well.
Energy harvesting from the sensor can provide this functionality. The most specific way to look at the
sensor as a means for energy harvesting is to model it as a capacitor. This model requires a dielectric
with a higher dielectric constant than the current sensor, but the idealized model will work for
discussion purposes. In the figure below, the energy generation component (to the far left) would be
the sensor.

Figure 127: Schematic of energy harvesting circuit courtesy of Los Alamos National Lab

As the sensor changes shape through motion on its application (such as a hydraulic hose on a piece of
machinery or a material transfer hose floating on the waves of the ocean) the capacitance of the sensor
would change. As the capacitance changes, the charge stored on the capacitor would be pushed
through the full wave rectifier and stored, temporarily, in the constant capacitor to charge a battery.
129

Over time, the battery would accumulate enough charge to check the status of the sensor and send out
a short message to the operators receiver to update the operator on the condition of the surface under
question.

8. Works Cited
1. Park, G., et al. Energy Harvesting for Structural Health Monitoring Sensor Networks. Los Alamos : Los
Alamos National Laboratory, 2007.
2. Yan, X., Want, Y., & Feng, X. Study for the enduracne of radial truck tires with finite element
modeling. Mathematics and computers in simulation. 2002, Vol. 59, 6.
3. Sundaramurthy, S. New Sensing Techniques for Structural Health Monitoring of Hydraulic Hose,
Composite Panels, and Biodegradable Metal Implants. 2011.
4. Andrews, Tobolsky. Stress Relaxation Studies of the Viscoelastic Properties of Polymers. Polymer
Science. 1952, Vol. 27, 7.
5. Parker Hannifin. Parker Product Browser. Parker. [Online] Parker Hannifin. [Cited: March 19, 2012.]
http://www.parker.com/portal/site/PARKER/menuitem.de7b26ee6a659c147cf26710237ad1ca/?vgnext
oid=fcc9b5bbec622110VgnVCM10000032a71dacRCRD&vgnextfmt=EN&vgnextdiv=&vgnextcatid=5965&
vgnextcat=792ST&Wtky=HOSE.
6. United States Department of Labor. Occupational Health and Safty Administration. [Online] [Cited:
March 19, 2012.] http://www.osha.gov/pls/imis/accidentsearch.accident_detail?id=170764609.
7. Parker Hannifin. Guide For Hose Failure Analysis. Parker Industrial Hose. [Online] 2012.
8. Kleiner, Gary and Honig, Doug. Injection Injury: Planning and preventing. Parker Hannifin Safety
Solutions. [Online] April 2009. [Cited: March 20, 2012.]
http://www.parker.com/parkerimages/Parker.com/ParkerStore%20and%20HOSE%20DOCTOR/North%2
0America/Home/Avoiding%20Injection%20Injuries.pdf.
9. Casey, Brenden. Hydraulics Super Market. How to Solve and Prevent Hydraulic Problems. [Online]
2005. [Cited: March 20, 2012.]
http://www.hydraulicsupermarket.com/download/hyd_problem_solving.pdf.
10. Hoon Sohn, Charles R Farrar. Damage diagnosis using time series analysis of vibration signals. s.l. :
Smart Materials and Structures, 2001. Vol. 10, 3.

130

11. Tractor Supply Company. Hydraulic Hose, 1/2 in. 2-Wire, 84 in. L. Tractor Supply Co. [Online]
February 5, 2012. [Cited: February 5, 2012.] http://www.tractorsupply.com/hydraulic-hose-1-2-in-2wire-84-in-l-1819842.
12. Gates Corporation. Hydraulics 101: Introduction to Hydraulics. Gates. [Online] [Cited: February 5,
2012.] http://www.gates.com/common/downloads/files/Gates/autoEducation/428-7153.pdf.
13. Risheill, Bob. Quote on Forklift Machinery. West Chester, OH : ProLift Equipment, February 10, 2012.
14. Wiring Practices for Hazardous Classified Locations Instrumentation : ANSI / ISA-TR12.6. Americna
National Standards Institute. s.l. : ANSI, 1995. ISBN: 1556175450.
15. Chawla, Krishan K. Composite Materials. Kito : Springer, 1987. ISBN: 0-387-98409-7.
16. Parker Hannifin. Literature. www.parker.com. [Online] [Cited: July 16, 2012.]
http://www.parker.com/literature/Hose%20Products%20Division/Catalog%204400%20PDF%20Files/Sec
tion_A_Hose.pdf.
17. Near-IR Fiber-Optic Temperature Sensor. Lin, Jie and Brown, Chris. 1, s.l. : Applied Spectroscopy, Vol.
47.
18. Strain Measurements of Silicon Dioxide Microspecimens by Digital Imaging Processing. Sharpe,
Pulskamp, Gianola, Eberl, Polcawich, Thompson. s.l. : Experimental Mechanics, 2007.
19. Opsens Solutions. O&G Solutions. opSens. [Online] Opsens, 2010. [Cited: March 29, 2012.]
http://www.opsens.com/en/oilandgas/.

9. Appendix
9.1.Data Analysis Code
%
%
%
%
%

Charles Dandino
James J. Sullivan
Comp Design 2
Final Project
3/9/2012

close all
clear all
clc
% Import Data
data = xlsread('Test20120224_CompDesign.xls','A4:D109');
% Locate data points that have no value and assign a vaule
loc_NaN = find(isnan(data(:,4)));
for v = 1:length(loc_NaN)
data(loc_NaN(v),4) = max(data(:,4));
end

131

% Define output: Resistance of Sameples (ohm)


y = data(:,4);
% Define factors
for ii = 1:size(data,2)-1
X{:,ii} = data(:,ii);
end
a = 'Strain'; % 0, 10% (tension) [X{1}]
b = 'Matrix Material'; % Liquid Paint (1), Rubber Cement (2) [X{2}]
c = 'Graphite Content (% by volume)'; % 50, 100, 150, 200, 250 [X{3}]
% ANOVA Table
[p,tbl,stats] = anovan(y,X,'model','interaction','varnames',{'A';'B';'C'},'display','off');
disp(tbl)
% Main effects plot
maineffectsplot(y,X,'varnames',{a;b;c});
% Interaction plot
figure
interactionplot(y,X,'varnames',{'A';'B';'C'});
% Store the unique values of each parameter
for ii = 1:size(X,2)
groupvals{ii} = unique(X{:,ii});
end
% Calcualte the residuals
index = 1;
residuals=[];
ybar=[];
for ii = 1:length(groupvals{1})
for jj = 1:length(groupvals{2})
for kk = 1:length(groupvals{3})
cell = data(find(data(:,1)==groupvals{1}(ii)),:);
cell = cell(find(cell(:,2)==groupvals{2}(jj)),:);
cell = cell(find(cell(:,3)==groupvals{3}(kk)),:);
ybar_temp = mean(cell(:,end),1);
residuals_temp = cell(:,end)-ybar_temp;
residuals = [residuals; residuals_temp];
ybar = [ybar ones(1,length(residuals_temp))*mean(cell(:,end),1)];
index = index+1;
end
end
end
% Sort and scale the residuals by the largest value for dimensionless
% residuals
residualsS = sort(residuals)/max(residuals);
figure
normplot(residualsS(3:(length(residualsS)-3)))
% Generate a liner regression w/ interaction
x0 = ones(length(y),1);
for ii = 1:size(X,2)
X_array(:,ii) = X{:,ii};
end
CONTs = regress(y,[x0 X_array X_array(:,1).*X_array(:,2) X_array(:,1).*X_array(:,3)]);

132

9.2.Data
Data taken on February 24, 2012
Strain Matrix Graphite Resistance over 2"

Sample File name


NA

test_001.lvm

NA

test_002.lvm

50

50

100

100

150

150

200

388110

200

197450

200

166530

200

48800

250

84130

250

390250

250

85770

39-1

test_004.lvm

250

97090

39-1

test_004.lvm

250

413820

39-1

test_005.lvm

250

89300

39-1

test_005.lvm

250

95370

39-2

test_006.lvm

250

92150

39-2

test_006.lvm

50

22960

39-3

test_007.lvm

50

24930

39-3

test_007.lvm

50

59710

40-1

test_008.lvm

133

50

70060

40-1

test_008.lvm

50

51400

40-2

test_009.lvm

50

128900

40-2

test_009.lvm

100

89300

40-3

test_010.lvm

100

235980

40-3

test_010.lvm

100

260540

40-3

test_011.lvm

100

136820

40-3

test_011.lvm

150

10520

40-3

test_012.lvm

150

29490

40-3

test_012.lvm

150

82030

40-3

test_013.lvm

150

33160

40-3

test_013.lvm

150

54650

40-3

test_014.lvm

150

25330

40-3

test_014.lvm

150

67390

40-3

test_015.lvm

150

56310

40-3

test_015.lvm

200

5660

46-1

test_016.lvm

200

25800

46-1

test_016.lvm

200

15570

46-2

test_017.lvm

200

12840

46-2

test_017.lvm

200

21060

46-2

test_018.lvm

200

16220

46-2

test_018.lvm

200

4860

46-2

test_019.lvm

200

11600

46-2

test_019.lvm

250

2940

46-3

test_020.lvm

250

13630

46-3

test_020.lvm

250

16040

46-3

test_021.lvm

250

36700

46-3

test_021.lvm

134

250

12520

47-1

test_022.lvm

250

12560

47-1

test_022.lvm

250

8500

47-2

test_023.lvm

250

12780

47-2

test_023.lvm

250

7760

47-2

test_024.lvm

10

50

47-2

test_024.lvm

10

50

47-3

test_025.lvm

10

100

47-3

test_025.lvm

10

100

48-1

test_026.lvm

10

150

48-1

test_026.lvm

10

150

48-1

test_027.lvm

10

200

413820

48-1

test_027.lvm

10

200

413750

48-1

test_028.lvm

10

200

413810

48-1

test_028.lvm

10

200

413820

48-2

test_029.lvm

10

250

221410

48-2

test_029.lvm

10

250

413820

48-2

test_030.lvm

10

250

413820

48-2

test_030.lvm

10

250

413480

48-3

test_031.lvm

10

250

413820

48-3

test_031.lvm

10

250

413820

48-3

test_032.lvm

10

250

413820

48-3

test_032.lvm

10

250

413820

48-3

test_033.lvm

10

50

281910

48-3

test_033.lvm

10

50

78540

49-1

test_034.lvm

10

50

104970

49-1

test_034.lvm

10

50

92820

49-1

test_035.lvm

135

10

50

209810

49-1

test_035.lvm

10

50

226290

49-1

test_036.lvm

10

100

44630

49-1

test_036.lvm

10

100

208050

49-2

test_037.lvm

10

100

141210

49-2

test_037.lvm

10

100

93710

49-2

test_038.lvm

10

150

179090

49-2

test_038.lvm

10

150

195370

49-2

test_039.lvm

10

150

190770

49-2

test_039.lvm

10

150

233360

49-3

test_040.lvm

10

150

217120

49-3

test_040.lvm

10

150

200800

49-3

test_041.lvm

10

150

190280

49-3

test_041.lvm

10

150

179000

50-1

test_042.lvm

10

200

100280

50-1

test_042.lvm

10

200

109380

50-1

test_043.lvm

10

200

107470

50-1

test_043.lvm

10

200

85330

50-1

test_044.lvm

10

200

82480

50-1

test_044.lvm

10

200

77890

50-2

test_045.lvm

10

200

68050

50-2

test_045.lvm

10

200

62180

50-2

test_046.lvm

10

250

39350

50-2

test_046.lvm

10

250

41600

50-2

test_047.lvm

10

250

37550

50-2

test_047.lvm

10

250

28110

50-3

test_048.lvm

10

250

34780

50-3

test_048.lvm

136

10

250

30830

50-3

Test_049.lvm

10

250

37710

50-3

Test_049.lvm

10

250

36230

50-3

test_050.lvm

10

250

34340

50-3

test_050.lvm

*Cells left blank are cells where the measurement was outside the range of the measurement
device.

9.3.Data Acquisition System

Sensor Skin Test Data Acquisition System


This document describes the elements and their configurations of data acquisition (DAQ) systems for
testing sensor skin, specifically tested in University of Cincinnati and conducted in Hose Product Division
of Parker Hannifin (Wickliffe, OH).

1. DAQ Module
Given sensor skin is designed to detect different failure modes based on the change of its electrical
resistance, NI 9219 is selected as it is a universal product that measures resistance, temperature, voltage
and current depending on how it is configured.

137

Figure 128 NI 9219 Channel and Terminal Assignments

Figure 129 NI 9219 Connections in 4-Wire Resistance and 4-Wire RTD Modes (reference: NI 9219 Operating Instructions and
Specifications)

To demonstrate sensor skin's capability to work with wireless system, NI WLS-9163 is selected as a
wireless carrier to connect with NI 9219 and transmit acquired signal through its wireless router to
designated laptop. Due to safety concern during lab tests, NI 9921 is used as Ingress Protection (IP)
enclosure to house 9219 and 9163. 9163 needs to be powered in order to transmit data. 9163 also
needs initial configuration on the laptop via Ethernet cable, however when it is configured it does not
need to connect the laptop through Ethernet any more.

138

Figure 130 Data Acquisition System

Picture above shows the setup of DAQ system, where the breadboard will be eventually enclosed and
wires will run through the I/O ports to connect with sensor skin electrodes.
The designated laptop is required to have National Instruments LabVIEW installed as the software to
communicate with DAQ module and configure data acquisition parameters (sampling rate, signal
manipulation, data storage etc.).

2. Measurement Circuit
9219 uses a 4-wire method to measure resistance, where its measurement range is 10.5k ohm. However
lab data show that actual data could exceed the range and reach mega ohm. Therefore, a parallel circuit
is used to decrease the range of overall resistance below 10.5k ohm (Figure 131) and a simple algorithm
in LabVIEW is deployed to compute the actual resistance (Equation 1) in real-time and store the
computed value.
Equation (1)

139

R1: 10k
R2: Sensor Skin

R: Measurement
Figure 131 Measurement Circuit for Using NI 9219

Resistors at different magnitudes are used to validate the circuit, where each of the resistors is
measured with a digital multi-meter (DMM) first and then plugged into the circuit to be measured by
the algorithm. Results are shown as below (the measured resistance of the other resistor is 10,152.1):
Table 1 Measurement Circuit Validation Results

DMM
Reading ()
2,185
2,178
9,890
9,970
21,880
21,700
32,470
32,600
149,200
149,400
1,010,000
1,010,000
10,520,000
10,120,000

Algorithm
Measurement ()
2,185.85
2,176.34
9,902.55
9,977.83
21,904.80
21,732.68
32,507.62
32,622.61
149,110.53
149,534.80
1,010,063.74
1,006,681.26
10,552,653.89
10,393,033.89

Error (%)
0.039038836
-0.076047914
0.126899725
0.078493577
0.113336662
0.150617533
0.115870343
0.069362267
-0.059966827
0.090226337
0.006311255
-0.328588404
0.310398183
2.697963329

Results show that the method has an acceptable measurement accuracy.

Appendix
1. Wireless settings to connect 9163 and laptop.

140

Figure 132 Laptop IP Configuration

Since the wireless connection is ad-hoc (point to point), the laptop needs to be under the same subnet
mask as the one of 9163. The laptop should connect to wireless network named WLS9163PH while 9163
is assured to be on and the "WLS LINK" LED light on 9163 is lit. Figure above shows the IP configuration
in order to connect successfully, while IP address of 9163 has already been set to 192.168.1.102 in NI
Measurement & Automation Explorer.

141

Você também pode gostar