Web PDF
Web PDF
SWINGLER
FOR THE Engineered Materials and Devices BEHAVIOR COLLECTION
ENGINEERING
Fundamentals and Principles
LIBRARY
Edited by Jonathan Swingler with major
Create your own
contribution by Alec Feinberg
Customized Content
Bundlethe more Degradation is apparent in all things and is fundamental to both
THE CONTENT
Manufacturing
key players engaged in the physics of degradation to present
the background science, current thinking and developments in
understanding, and gives a detailed account of emerging issues
Degradation in
Engineered Materials
across a selection of engineering applications. The work has been
Engineering
put together to equip the upper level undergraduate student,
Mechanical
postgraduate student, as well as the professional engineer and
& Chemical
and Devices
scientist, in the importance of physics of degradation. The aim
Engineering of The Physics of Degradation in Engineered Materials and
Materials Science Devices is to bridge the gap between published textbooks
& Engineering on the fundamental science of degradation phenomena
Civil & and published research on the engineering science of actual
Environmental
Engineering
fabricated materials and devices.
A history of the observation and understanding of physics of Fundamentals and Principles
Advanced Energy degradation is presented and the fundamentals and principles
Technologies of thermodynamics and entropy are extensively discussed. This
is the focus of this book, with an extended chapter by Alec
Feinberg on equilibrium thermodynamic damage and non-
THE TERMS equilibrium thermodynamic damage. It concludes with two
Perpetual access for particular technologies to give examples of areas of application.
a one time fee
No subscriptions or Dr. Jonathan Swingler received a Joint BSc (Hons) in physics
access fees and chemistry from Keele University in 1990, followed by a PhD
at Loughborough University for his work on the degradation of
Unlimited
electrical contacts. He subsequently moved to the University of
concurrent usage
Southampton where he pursued his research into the physics of
Downloadable PDFs
degradation and reliability of electrical and electronic materials
Free MARC records and devices. Currently Jonathan is working to develop reliability-
engineering science at Heriot-Watt University, to enable the
Edited by Jonathan Swingler
For further information,
a free trial, or to order,
contact:
prediction of electrical component and system reliability under
harsh operating conditions. He is an associate professor of with major contribution
energy at Heriot-Watt University and fellow of the Institute of
sales@momentumpress.net Physics. by Alec Feinberg
ISBN: 978-1-60650-467-3
THE PHYSICS OF
DEGRADATION
IN ENGINEERED
MATERIALS AND
DEVICES
THE PHYSICS OF
DEGRADATION
IN ENGINEERED
MATERIALS AND
DEVICES
Fundamentals and Principles
DOI: 10.5643/9781606504680
10 9 8 7 6 5 4 3 2 1
KEY WORDS
List of Figures xi
List of Tables xv
Chapter 1Introduction 1
1.1Introduction 1
1.2 Rationale and Emphasis 1
1.3 Thermodynamics and Entropy 3
1.4 Technologies and Applications 7
1.5Conclusion 9
References 10
Chapter 2History of the Physics of Degradation 11
2.1Context 11
2.2History 12
2.3Conclusion 18
References 19
Chapter 3Thermodynamics of Ageing and Degradation
inEngineering Devices and Machines 23
3.1Introduction to Degradation and Ageing 23
3.2Thermodynamic Degradation Paradigm 24
3.3 Review of the DEG Theorem 24
3.4 Review of Thermodynamics 27
3.5Entropy and Production of Irreversible Entropy 28
3.6 Dissipative Mechanisms and Ageing 30
3.7Example Applications of the DEG Theorem 34
3.8Conclusion 45
References 45
viii Contents
6.2Humidity 155
6.3Temperature Effects on Reliability
of Piezoelectric Actuators 164
6.4 Measurement Techniques 173
6.5Conclusion 176
Acknowledgments 177
References 177
About the Contributing Authors 181
Index 183
List of Figures
at 10 kA/Cm2.109
Figure 4.9. Life test data of gate-source MESFET leakage current
over time fitted to the ln(1 + Bt) ageing model.
Temperatures are ambient. Junction rise is about 30C
(Feinberg et al. 2000). 111
Figure 4.10. Ageing over all time t with critical values tc prior to
catastrophic failure 3. 114
xii List of Figures
Introduction
Jonathan Swingler
1.1INTRODUCTION
This concise reference work has been put together to equip upper-level
undergraduate and postgraduate students as well as professional engineers
and scientists with the importance of degradation in engineered materials
and devices. It brings together experts and key players engaged in the
physics of degradation to present the background science, current think-
ing, and developments in understanding and give a detailed account of the
emerging issues across a selection of engineering applications.
This introductory chapter presents the rationale and emphasis of the
book. It explains the relevance of what is termed in the title and the sub-
title of the work. The chapters on thermodynamics and entropy are very
briefly surveyed, along with two other chapters on example technologies
and applications to show the significance of the subject area.
The aim of the work is to bridge the gap between published textbooks
on the fundamental science of degradation phenomena and published
research on the engineering science of actual fabricated materials and
devices.
DOI: 10.5643/9781606504673/ch1
2THE PHYSICS OF DEGRADATION
Corrosion
Digestion
Diffusion
Adhesion
Reaction
Displace
Deform
Facture
Mixing
Creep
Wear
Ogejcpkuou
Fkuekrnkpg"ecvgiqtkgu
Physics of degradation
The title contains the phrase engineered materials and devices, par-
ticularly to emphasize engineered as the designer or architect is inter-
esting in developing or fabricating a material or device that has a specific
purpose or function. The scope of the materials and devices themselves
span a vast array of products from mechanical to electrical, to microelec-
tronic, and so on.
The target readers are those who are interested in taking thermodynamics
of ageing beyond the theoretical ideas to specific applications. For the
upper-level undergraduate student embarking on project work, it opens
up opportunities to develop predictions from theory and conduct experi-
ments to explore the limits of those predictions. For the postgraduate and
postdoctoral researcher, it offers springboard ideas to push the boundaries
of current thinking. For the professional engineer and scientist, it gives
grounding in the current thinking of the underlying science to enable the
development of innovative solutions.
The contributing authors are scientists and engineering working in
the field from both academia and industry. There is a wealth of experience
from initiating novel concepts for pushing forward the research effort,
developing innovative solutions for commercial products, and communi-
cating these ideas.
The DEG theorem is presented by Bryant in Chapter 3. This links the rate
of degradation or ageing to rates of production of irreversible entropy by
underlying dissipative physical processes. The second law of thermody-
namics plays a central role here. To explain this approach, a review of
thermodynamics is given in the chapter discussing entropy and the pro-
duction of irreversible entropy under particular conditions. Four condi-
tions are identified:
The chapter gives three detailed examples of using the DEG theorem
to describe degradation due to:
approach. The system degradation path to a minimum may not be the final
minimum available to the system. An energy barrier or barriers may exist
to further or the final minimum.
1.4.2PHYSICS OF DEGRADATION IN
FERROELECTRIC DEVICES
Only two applications are presented in this volume to illustrate the signifi-
cance of the physics of degradation to commercial products. The electrical
power system application could be argued is well established in reliability
issues and ageing, whereas many aspects of ferroelectric technologies are
still an emerging area. Concerning reliability in electrical and electronic
systems, the reader is directed to texts like Reliability Characterisation
of Electrical and Electronic Systems edited by Swingler (2015). There
are many more established and emerging technologies that are at various
stages of maturity in having a coherent understanding of the ageing phe-
nomena and reliability. These are shown in Figure 1.2, which is not an
Introduction 9
Molecular electronics
Polymer electronics
Electromechanical
Microelectronics
Photoelectronics
Bio-Electronics
Ferroelectric
Wearable
Electrical
MEMS
NEMS
Mechanical Thermal Chemical Electrical Biological
1.5CONCLUSION
This introductory chapter has laid out the aim of the book. The book is
designed to bridge the gap between published textbooks on the funda-
mental science of degradation phenomena and published research on the
engineering science of actual fabricated materials and devices. The ratio-
nale and emphasis of the book are presented, explaining the relevance of
the subject matter.
An introduction to the physics of degradation is given with an illustra-
tion of the types of mechanisms involved. These mechanisms cover a wide
range of disciplines such as thermal, mechanical, chemical, biological, and
so on and are underpinned by thermodynamics and entropy. The chapters
on thermodynamics and entropy are very briefly surveyed along with the
two chapters on example technologies and applications to show the signif-
icance of the subject area. Further established and emerging technologies
are illustrated, showing the wide scope of the applicable area. These are at
different levels of maturity in having a more complete understanding of the
degradation phenomena leaving much room for research and innovation.
10THE PHYSICS OF DEGRADATION
REFERENCES
Mandl, F. 1989. Statistical Physics. 2nd ed. Chichester, UK: John Wiley and Sons.
Swingler, J. 2015. Reliability Characterisation of Electrical & Electronic Systems.
Cambridge, UK: Woodhead Publishing, an imprint of Elsevier.
CHAPTER 2
Jeff Jones
2.1CONTEXT
DOI: 10.5643/9781606504673/ch2
12THE PHYSICS OF DEGRADATION
the ability to forecast how the parameter changes with time. This also
implies the use of a model of some kind. It is not possible to think about
degradation without considering a model, which means, the physical pro-
cess of degradation and its modelling cannot be treated separately.
Therefore, in any history of system degradation, it is not possible to
identify where the science of degradation began. Similarly, we cannot
trace the history up to the present time, since, in this area (which now
forms part of asset management) research is moving along at a pace, espe-
cially in the prognostics field, where anything that is written would be
outdated in a very short time.
This leads to a dilemma about what exactly should be in a document
about the history of degradation? The approach taken here has been to try
to identify key features that have led to where we are today. To do this
limits how far we can look back, since it is necessary to have evidence
and knowledge, which can only be achieved through the published work
in this area. Since we are limited to approximately the past 100 years. the
real title of this chapter should be Key Moments in the History of Degra-
dation Over the Past 100 or So Years.
2.2HISTORY
This paper suggests that by this time it is known that fatigue can
cause failure, which is already a well-known concept, but that another
process, in this case, oxidation, changes the material properties such that
the fatigue life is changed. It is the start of this sort of understating that
the system mechanical properties are changed by environmental, in this
case, chemical, effects that leads to the understanding of the physics of
the failure itself.
The first proper systematic study of fatigue life that is recorded is that
of Whler, who around 1867 published work at the Paris exposition that
examined the fatigue behavior of materials (Whler 1870a). Of course,
earlier, he too had been examining railway axles, from 1858 to 1860, and
had been developing deflection gauges (Whler 1858, 1860) to measure
them under load.
His work at that time was driven by safety issues. There had been
a number of accidents in the railway industry (Smith 1990), and he was
attempting to work out the probability of failure for different designs and
so develop a safety factor that could be used to make designs safer. In
fact, his work touched on many areas, the measurement of service loads
and the calculation of the corresponding stresses, the design for life and
estimation of the probability of survival, as well as some observation on
crack propagation (Whler 1870b).
Whlers work is the first exploration of what were to be known
as SN curves, or alternatively, Whler curves, which characterize the
fatigue behavior of materials. Whler himself drew no curves, present-
ing his data as tables, and the curves themselves were first drawn by
his successor Spangenberg (1874, 1875, 1879). The curves show stress
applied on the vertical axis, with a number of application or cycles
before failure on the horizontal axis, as in Figure 2.1. The lines rep-
resent the behavior of the material and are different for each material.
The endurance limit, which is sometimes known as the fatigue limit or
14THE PHYSICS OF DEGRADATION
U/P"Ewtxg
50
45
40
Steel
35
Endurance limit
30
Uvtguu"*muk+
Aluminium
25
20
15
10
5
0
103 104 105 106 107 108 109
Pwodgt"qh"e{engu
In this work, Austin identified the likely cause of the degradation and
noted that developing the correct tests for degradation was important. This
is probably the first reference to testing to understand the rate of degrada-
tion under different environmental conditions.
The data gathered by Whler (1870a) was analyzed by Basquin in
1910 using slightly more up-to-date mathematical analysis techniques to
produce the initial loglog models for stress life (Basquin 1910). Further,
in a similar vein Miner (1945) formalized the linear damage hypothesis
that was first suggested by Palmgren (1924). In addition to the Palmgren
Miner rule, this paper contained a four-parameter equation relating the
tensile strength to the fatigue limit for the SN curve, a relationship that
was later used by Weibull. Palmgren (1924) also specified B10-life as a
design criterion, which is a survival probability statement and the first case
was developed for ball bearings.
The PalmgrenMiner rule, more commonly known only as Miners
(1945) rule, is an early example of a damage accumulation model that
deals with failure caused by fatigue. This is a progressive, localized, per-
manent damage that occurs when a material is subject to cyclic stresses
below the tensile strength limit of the material. Failure actually begins with
the initiation of a microscopic crack in the material, which then develops
in a fairly predictable way. This predictability arises since the actual crack
propagation rate is a property of the material and is related to the applied
stress level and the amount of damage that already exists in the material.
Eventually, the crack will become unstable under the applied stress and
failure will occur. This will happen after a predictable duration for any
particular material and stress level. It is this predictability that leads to
degradation models such as the Whler or SN curves (Whler 1870a),
Miners law (Miner 1945), Pariss relationship (Paris 1961, 1964), and the
CoffinManson relation (Coffin and Manson 1954).
Paris (1961) had established that fatiguecrack propagation could be
described by a simple equation that can be considered equivalent to an
SN curve under certain limiting conditions and with many assumptions.
The model was later refined by Paris (1964) to remove some limitations
and generalize it. For a while it was considered to be a large step for-
ward to be able to calculate the propagation of cracks of arbitrary shape
under arbitrary stress types even though the equation and some of its
assumptions are now considered to have vastly oversimplified the phys-
ical reality.
In 1954, and based on the work by Bauschinger (1880, 1881, 1886),
Coffin (1954a, 1954b) and Manson (1953) described the behavior of metal-
lic materials under cyclic inelastic strain amplitudes using a four-parameter
16THE PHYSICS OF DEGRADATION
relationship that bears their name and creates a new field, known as low-
cycle fatigue analysis, which is applied to items loaded over relatively
fewer cycles at elevated temperatures.
Fatigue is in fact one of the most persistent problems in engineer-
ing design, causing failure in everything, from rotating and reciprocating
machinery and components, through failure in complex systems such as
aircraft and ships, to large civil engineering structures such as bridges and
buildings. Even at the electronic component level, fatigue is a major prob-
lem in solder joints, bond wires, and copper-plated vias (Dasgupta 1993).
In 1961, Professor Waloddi Weibull published a book on materials and
fatigue testing while working as a consultant for the U.S. Air Force Mate-
rials Laboratory (Weibull 1961). Prior to this work, Weibull had originated
one of the most useful techniques aiding people in understanding and
analyzing degradation, which originated from a paper published in 1951
(Weibull 1951). Up until about 1963, Weibull, who was then a Swedish
mechanical engineer, also wrote a number of papers related to fatigue
and other failure mechanisms. Weibulls original paper (Weibull 1951)
did little more than propose a multiparameter distribution, but it became
widely appreciated and was shown by Abernethy, working with Pratt and
Whitney in 1967, to have some application to the analysis of defect data
(Pratt and Whitney 1967). The versatility of this technique allows it to
mimic a number of other distributions, including the log-normal dis-
tribution, which is one of the main distributions used for the analysis of
degradation. This can be theoretically derived under assumptions match-
ing many degradation processes common to material failure mechanisms
such as corrosion, diffusion, migration, crack growth, electromigration,
and many other chemical reactions and processes (Epstein 1947). While
the Weibull distribution has proven to be very powerful and extremely
useful in this area, it should be noted that Lawless (1983) commented that
an enormous amount of detailed work on the Weibull distribution should
have been directed to the investigation of other statistical models.
The log-normal distribution is often considered for modelling the
failure time distribution when the corresponding degradation process is
based on rates that combine multiplicatively, and is convenient for mod-
elling fatigue crack growth in metals and composites as proposed by Kao
(1965). The log-normal distribution is an exponential transformation of
the normal distribution and fits many diverse types of data well because it
has a great variety of shapes. It is widely used for life data showing early
increasing failure rates, and later, decreasing failure rates with increasing
age (Lamberson and Kapur 1977; Nelson 1990; Yang 2007). Examples
of its application include metal fatigue, electronic components, electrical
History of the Physics of Degradation 17
causes failure. This was an important step forward for new models that
described degradation processes. Meeker and Escobar, much later in 1998,
showed that most degradation models will not yield lifetime distributions
with closed-form solutions (Meeker and Escobar 1998).
One of the main uses of degradation modeling has been to use it
in order to provide an input for reliability prediction. Gertsbakh (1989)
suggests special attention to lifetime investigation for failure caused by
fatigue. Lu and Meeker (1993) combined regression model, simulation,
and bootstrap methods and applied it to degradation data in order to esti-
mate the time-to-failure distribution. Tang and Chang (1995) described
a conceptual framework for reliability evaluation from non-destructive
accelerated degradation data, where the data were treated as a collection
of stochastic processes for which parameters depend on stress level. Yang
and Yang (1998) developed a method for rapidly evaluating reliability
using degradation data. Crk later in 2000 extended Lu and Meekers work
in developing a multiple multivariate regression model that considered the
case when the parameters are random, dependent, and vary with applied
stresses (Crk 2000).
Degradation models are getting increasingly important for modern
high-cost and high-reliability products, such as aerospace components,
where generating failure data through life testing is time consuming and
costly. Here, the use of degradation models can be used to shorten testing
since time to failure can be predicted when enough data has been collected
(Jones 1999). It is also possible, if failure is defined in terms of a specified
level of degradation, to use a degradation model to define a particular
time-to-failure distribution (Amster and Hooper 1983; Lu and Meeker
1993; Meeker and Escobar 1998; Nelson 1990).
Understanding of degradation processes and degradation modelling
has evolved since the early days in the 1830s and the modern engineer or
scientist has a vast array of tools and techniques to choose from to help
understand the degradation process in the numerous materials used today
in technological systems.
2.3CONCLUSION
The material that has been covered in this chapter is selective and is what the
author feels are the most important publications and by implication, pieces
of work, in the physics of degradation field over the past 100 years. An
attempt has been made to generalize and to look at publications that address
more general themes rather than the detailed history of the d evelopment of
History of the Physics of Degradation 19
REFERENCES
Aitchison, J., and J.A.C. Brown. 1957. The Log Normal Distribution. New York
and London: Cambridge university press.
Albert, W.A.J. 1837. liber Treibseile am Harz. Archiv fur Mineralogie, Geor-
gnosie. Bergbau und Huttenkunde 10, pp. 21534.
Amster, S.J., and J.H. Hooper. 1983. Accelerated life tests with measured degrada-
tion data and growth curve models. American Statistical Association Annual
Meeting, Vol. 5, no. 9.
Austin, A.O. December 11, 1914. Insulator Depreciation and Effect on Operation
302d Meeting of the American Institute of Electrical Engineers, New York.
Basquin, O.H. 1910. The exponential law of endurance tests. Proceedings of
American Society of Testing Materials, Vol. 10, pp. 62530.
Bauschinger, J. 1880. Uber das Kristallinischwerden und die Festigkeitsvermin-
derung des Eisens durch den Gebrauch. Dinglers J. 235, pp. 16973.
Bauschinger, J. 1881. Uber die Veranderung der Elastizitatsgrenze und des
Elastizitatsmoduls verschiedener Metalle. Civilingenieur 27, pp. 289348,
Leipzig: Felix.
Bauschinger, J. 1886. Uber die Veranderung der Elastizitatsgrenze und der Fes-
tigkeit des Eisens und Stahls durch Strecken und Quetschen, durch Erwarmen
und Abkiihlen und durch oftmals wiederholte Beanspruchung. Mitt. Mech.
Tech. Lab. Munch. 13, pp. 1115.
Beau, J.F. 1964. Management of the human element in the physics of failure. Pro-
ceedings of 3rd Annual Symposium on the Physics of Failure in Electronics.
pp. 26479.
Birnbaum, Z.W., and S.C. Saunders. 1969. A new family of life distributions.
Journal of Applied Probability 6, no. 2, pp. 31927.
Braithwaite, F. 1854. On the fatigue and consequent fracture of metals. Institu-
tion of Civil Engineers, Minutes of Proceedings, Vol. 8, pp. 46374. London.
Coffin, L.F. 1954a. A study on the effect of cyclic thermal stresses on a ductile
metal. Transaction of American Society of Mechanical Engineers (ASME)
76, pp. 93150.
20THE PHYSICS OF DEGRADATION
Coffin, L.F. 1954b. The problem of thermal stress fatigue in austenitic steels at
elevated temperatures. ASTM STP No. 165, American Society for Testing
and Materials, Symposium on Effect of Cyclic Heating and Stessing on Met-
als at Elevated T emperatures.
Crk, V. 2000. Reliability assessment from degradation data. Proceedings Annual
Reliability and Maintainability Symposium (RAMS), pp. 15561.
Crook, D.L. 1979. Method of determining reliability screens for time dependent
dielectric breakdown. Proceedings of 17th Annual Reliability Physics Sympo-
sium, pp. 17.
Darrow, K.K. June, 1942. Entropy. Bell System Technical Journal 21, no. 1, pp.
51, 74, doi: 10.1002/j.1538-7305.1942.tb01301.x
Dasgupta, A. 1993. Failure mechanism models for cyclic fatigue. IEEE Trans-
actions on Reliability 42, no. 4, pp. 54855.
Davis, H. September 2627, 1962. Introduction. Proceedings of the 1st Annual
Symposium on Physics of Failure in Electronics, pp. 13.
Dodson, G.A., and B.T. Howard. January, 1961. High stress aging to failure of
semiconductor devices. Proceedings of 7th National Symposium on Reliabil-
ity and Quality Control. Philadelphia, PA.
Epstein, B. December, 1947. The Mathematical Description of Certain Breakage
Mechanisms Leading to the Logarithmic-Normal Distribution. Journal of
the Franklin Institute 244, no. 6, pp. 47177.
Gertsbakh, I.B. 1989. Statistical Reliability Theory. New York: Marcel Dekker
Inc, pp. 427.
Gertsbakh, I.B., and K.B. Kordonsky. 1969. Models of Failure. Berlin: Springer.
Graves, J. 1884. On the Causes of Failure of Deep-Sea Cables. Journal of the
Society of Telegraph-Engineers and Electricians 8, no. 51, pp. 11926.
Hieber, H., and K. Pape. 1984. Lifetime of bonded contacts on thin film met-
allization. Proceedings of 22nd Annual Reliability Physics Symposium, pp.
12833.
Hoyland, A., and M. Rausand. 1994. System Reliability Theory, Models and Statis-
tical Methods. New York: John Wiley & Sons.
Jones, J.A. 1999. A Toolkit for Parametric Drift Modelling of Electronic Compo-
nents. Reliability Engineering and System Safety 63, pp. 99106.
Kao, J.H.K. 1965. Statistical Models in Mechanical Reliability. Proceedings
of the 11th National Symposium on Reliability and Quality Control. pp.
2407.
Kapur, K.C., and L.R. Lamberson. 1977. Reliability in Engineering Design. New
York: John Wiley & Sons.
Lawless, J.F. 1983. Statistical Methods in Reliability. Technometrics 25, no. 4,
pp. 30516.
Lu, C.J., and W.Q. Meeker. 1993. Using Degradation Measures to Estimate a
Time-to-failure Distribution. Technometrics 35, no. 2, pp. 16173.
Manson, S.S. 1953. Behaviour of materials under conditions of thermal stress.
Heat Transfer Symposium, University of Michigan Engineering Research
Institute, pp. 975.
History of the Physics of Degradation 21
Meeker, W.Q., and L.A. Escobar. 1998. Statistical Methods for Reliability Data.
John Wiley & Sons.
Miner, M.A. September, 1945. Cumulative Damage in Fatigue, Journal of
Applied Mechanics 12, no. 3, pp. A l5964.
Morin, A. 1853. Lecons de mecanique practiqueresistance des materiaux. Paris,
Librairie de L. Hachette et Cie, p. 456.
Nelson, W. 1990. Accelerated Testing: Statistical Models, Test Plans and Data
Analysis. New York: John Wiley & Sons.
Palmgren, A. 1924. Durability of Ball Bearings. ZDVDI 68, no. 14, pp. 339,
(published in German).
Paris, P.C. 1964. The Fracture Mechanics Approach to Fatigue, In FATIGUE
An Interdisciplinary Approach, eds. J.H. Burke, N.L. Reed, and V. Weiss.
New York: Syracuse University Press.
Paris, P.C., M.P. Gomez, and W.E. Anderson. 1961. A Rational Analytic Theory
of Fatigue. The Trend in Engineering, 13, pp. 914.
Place, C.S., J.E. Strutt, K. Allsopp, P.E. Irving, and C. Trille. 1999. Reliability
Prediction of Helicopter Transmission System Using Stress-strength Interfer-
ence with Underlying Damage Accumulation Quality and Reliability Engi-
neering International, 15, pp. 6978.
Pratt and Whitney Aircraft. 1967. Introduction to Weibull Analysis. PWA 3001:
East Hartford, CT.
Rankine, W.J.M. 1842. On the Causes of the Unexpected Breakage of the Jour-
nals of Railway Axles, and on the Means of Preventing Such Accidents by
Observing the Law of Continuity in Their Construction. Institution of Civil
Engineers, Minutes of Proceedings, Vol. 2, pp. 1058. London.
Shiomi, H. 1965. Cumulative degradation model and its application to component
life estimation. Proceedings of Fourth Annual Symposium on the Physics of
Failure in Electronics. pp. 7494.
Smith, R.A. 1990. The Versailles railway accident of 1842 and the first research
into metal fatigue. In vol. 4 of Fatigue 90, Birmingham: EMAS.
Sobczyk, K., and B.F. Spencer. 1992. Random Fatigue from Data to Theory.
Boston: Academic Press.
Spangenberg, L. 1874. Uber das Verhalten der Metalle bei wiederholten Anstren-
gungen. Z. Bauw. 24, pp. 47395 and 25, pp. 7898.
Spangenberg, L. 1875. Uber das Verhalten der Metalle bei wiederholten Anstren-
gungen. Berlin, Ernst and Korn.
Spangenberg, L. 1879. Uber Festigkeits-Versuche mit Eisen und Stahl. Glasers
Ann. Gew. 5, pp. 615.
Tang, L.C., and D.S. Chang. 1995. Reliability Prediction Using Non-Destructive
Accelerated Degradation Data: Case Study on Power Supplies. IEEE Trans-
action on Reliability 44, no. 4, pp. 5626.
Weibull, W. 1961. Fatigue Testing and Analysis of Results. London: Pergamon
Press.
Weibull, W. November, 1951. A Statistical Distribution Function of Wide Appli-
cation. Journal of Applied Mechanics 18, pp. 2937.
22THE PHYSICS OF DEGRADATION
Whler, A. 1858. Bericht iiber die Versuche, welche auf der konigl. Nieder-
schlesisch- markischen eisenbahn mit Apparaten zum Messen der Biegung
und Verdehung von Eisenbahnwagenachsen wahrend der Fahrt angestellt
wurden. Zeitschrift fur Bauwesen 8, pp. 64152.
Whler, A. 1860. Versuche zur Ermittlung der auf die Eisenbahnwagenachsen
einwirkenden Krafte und die Widerstandsfahigkeit der Wagen-Achsen.
Zeitschrift fur Bauwesen 10, pp. 583616.
Whler, A. 1870a. Uber die Festigkeitsversuche mit Eisen und Stahl. Auf Anord-
nung des Ministers fur Handel, Gewerbe u. offentl. Arbeiten, Grafen Itzen-
blitz, angestellt. Berlin: Ernst und Korn.
Whler, A. 1870b. ber die Festigkeitsversuche mit Eisen und Stahl. Zeitschrift
fr Bauwesen 20, pp. 73106.
Wolstenholme, L.C. 1999. Reliability Modelling: A Statistical Approach. New
York: Chapman and Hall/CRC.
Yang, G. 2007. Life Cycle Reliability Engineering. New York: John Wiley & Sons.
Yang, K., and G. Yang. 1998. Degradation Reliability Assessment Using Severe
Critical Values. International Journal of Reliability, Quality and Safety
Engineering 5, no. 3, pp. 8595.
York, J.O. 1842. An Account of a Series of Experiments on the Comparative
Strength of Solid and Hollow Axles. Institution of Civil Engineers, Minutes
of Proceedings, Vol. 2, pp. 1058, London.
CHAPTER 3
Thermodynamics of
Ageing and Degradation
inEngineering Devices
and Machines
Michael D. Bryant
3.1INTRODUCTION TO DEGRADATION
AND AGEING
DOI: 10.5643/9781606504673/ch3
24THE PHYSICS OF DEGRADATION
3.2THERMODYNAMIC DEGRADATION
PARADIGM
{ ( )}
w = w pi z ij , (3.1)
( )
which depends monotonically on the energies pi z ij associated with the
i = 1, 2,n underlying irreversible dissipative processes, where index j
refers to the time-dependent phenomenological variables z ij that charac-
terize each process energy pi. To be in accordance with the second law
of thermodynamics (Kondepudi and Prigogine 1998), the dissipative irre-
versible processes must produce irreversible entropy:
{ ( )} , (3.2)
S = S pi z ij
( )
that must depend on the same pi z ij . Since the second law demands dS 0,
the function in Equation 3.2 must be monotonic and positive. The apostrophe
notation indicates irreversible entropy. Through the second law of thermody-
namics, the structure of dependency of S on the phenomenological variables
z ij is similar to that of w. The DEG theorem exploits this dependency. By
applying the chain rule to Equations 3.1 and 3.2, rates of degradation and
entropy are
dw w pi z ij dS ' S ' pi z ij
= j
,
dt
= j
i , j pi z i t
. (3.3)
dt i , j pi z i t
dw w / pi S ' pi z ij S ' pi z ij
= j
= Bi .(3.4)
dt i , j S '/ pi pi z i t i, j pi z ij t
26THE PHYSICS OF DEGRADATION
w (3.5a)
w / pi S ' pi
Bi = =
S '/ pi w
(3.5b) T
i pi
that arises from the term within the curly brackets in Equation 3.4. In
Equation 3.5a, the subscript pi denotes w/S with dissipative process pi
active, or geometrically, w/S along the pi direction in relevant multi-
dimensional spaces that graph Equations 3.1 and 3.2 as surfaces versus
coordinates pi. The Bi, material properties that can be measured or related
to other material properties, may be constant or depend on the thermo-
dynamic states such as temperature. Definition of entropy (and the units
involved) suggests S/pi = 1/Ti, where Ti is a temperature associated with
process energy pi. With this, Equation 3.5b gives another way to obtain
the Bi. The terms in Equation 3.4 that multiply S/pi involve dissipated
power components
dpi p z j
= j ij i = j eij f ij (3.6)
dt z i t
associated with the ith dissipative process. The products within the
sums are power conjugates involving a rate of a phenomenological vari-
able z ij/t analogous to a generalized velocity fij, and its power conjugate
pi /z ij analogous to a generalized force eij. Equation 3.4 becomes
dw S ' pi z ij eij f ij
= Bi j
= Bi
dt i, j pi z i t i, j Ti
(3.7)
dp / dt dS 'i
= Bi i = Bi
i Ti i, j dt
The last two terms in Equation 3.7 explicitly show degradation rate
to be a linear combination of the rates of entropy production (or power
dissipation when relevant), where Si and dpi /dt are entropy generated and
power dissipated by the ith dissipative process. If entropy or power dissi-
pated can be formulated, a quantitative degradation model can be formu-
lated that is underpinned by the laws of thermodynamics. Thermodynamics
THERMODYNAMICS OF AGEING AND DEGRADATION 27
All systems must conserve energy, stated by the first law of thermodynamics
dE = dQ dW + k hk dN k . (3.8)
E
hk = (3.9)
N k
consists of a reversible change dSe from flow of entropy into and out
of the control volume, and an irreversible change that arises from gen-
eration of entropy dS inside the control volume. Entropy flow dSe arises
from heat transfer
TdSe = dQ + k hk d e N k . (3.11)
due to flow of heat dQ (via conduction and radiation) and flow of matter
deNk across the control volume boundary. Here, chemical potential hk
gauges the energy contained per mole of matter. Change in number of
moles
dN k = d ' N k + d e N k (3.12)
consists of a reaction term dNk (chemical or phase change) and the matter
transport term deNk. Note the terms deNk extend the foregoing equations to
open systems.
28THE PHYSICS OF DEGRADATION
dE dW 1
dS ' = dS + k hk d ' N k . (3.13)
T T T
SdT dA dW 1
dS ' = + k hk d ' N k, (3.15)
T T T T
THERMODYNAMICS OF AGEING AND DEGRADATION 29
Vdp dH dW ' 1
dS ' = dS + + k hk d ' N k , (3.16)
T T T T
which needs conditions on dG, dp, and dT. In Equations 3.16 and 3.17,
dW = dW - pdV denotes work other than pressure volume.
Stationary systems: The simplest condition for Equation 3.13 is the sta-
tionary system with dE = dS = 0 (constant internal energy and entropy).
Here, combinations of flows across, and production inside the control vol-
ume of energy and entropy balance, rendering constant levels of internal
energy and entropy within the control volume. Stationary systems are at a
steady state but not in equilibrium.
Systems under strain and isothermal: Conditions for Equation 3.15 involve
dA and dT. Helmholtz free energy A, energy untainted by entropy or exergy,
can include energy associated with conservative forces (such as elastic
strain energy of deformed material dependent on displacements), and the
work of isothermal separation of surfaces, important for quantifying frac-
ture and fatigue phenomena, among others. For isothermal systems, dT = 0.
Isentropic and isobaric systems: Conditions for Equation 3.16 involve dH,
dS, and dp. Enthalpy H, with pressure volume expansion work subtracted
from internal energy, pertains to gases and phase changes of materials. For
isothermal systems, dp = 0.
Isothermal and isobaric systems: Conditions for Equation 3.17 involve dH,
dT, and dp. Gibbs free energy G, with pressure volume expansion work
subtracted from internal energy, pertains to dT = dp = 0 (constant tempera-
ture and pressure) conditions under which chemical reactions are standard-
ized. For isothermal systems, dp = 0. This dissipated work must eventually
diffuse through heat flow dQ or mass flow S hk dNk or both, as entropy flow
dSe. Open systems demand balancing flows of entropy, heat, work, energy,
and mass over a control volume about the degrading body or system.
30THE PHYSICS OF DEGRADATION
g
S ' = As , (3.18)
Tm
where the interface surface energy Dg is the work per unit area needed to
separate surfaces in a solid to create new surface area DAs, and Tm is the local
temperature of the media (Maugis 1999). Surface energies g are 1 to 3 Jm2
for clean metals, 0.1 to 0.5 Jm2 for ionic crystal ceramics, and less than
0.1Jm2 for molecular crystals and polymers (Maugis 1999).
Plastic Deformation, Viscous Dissipation, and Cutting of solids induce
an entropy change
Uc
S ' = V , (3.19a)
Tm
U c = sd e . (3.19b)
Here s is the stress tensor and e is the strain tensor. The form of Equa-
tion 3.19b is valid for solids and fluids. For plastic dissipation in solids, s
is the deviatoric stress tensor.
Fracture separates continuous materials and forms new free surfaces
within the cracked body. Overstressed bodies fracture, especially brittle
materials. Fracture is associated with rupture of bodies, fatigue of bodies,
and surface damage. For a crack of length a, in a linearly elastic brittle
material, under plane strain, at uniform temperature Tcr, with no material
THERMODYNAMICS OF AGEING AND DEGRADATION 31
loss, and no chemical reactions or phase changes, Rice (1978) derived the
irreversible entropy produced
G 2go
dS ' = da , (3.20)
Tcr
where energy release rate G = -Us/a is the energy released from the
elastic strain energy Us per increment of crack extension and go is the sur-
face energy needed to establish free surfaces in the solid under isothermal
conditions.
Phase changes, including melting, solidification, crystallization, and
evaporation, alter material structure. Temperature ageing can induce a
phase change. Material properties such as strength or elastic modulus of
components can degrade. Phase changes, associated with surface melting
and recrystallization of metals, induce entropy change
H
S ' = (3.21)
Tphase
reactants
ki ri
products
ki pi . (3.22a)
A
dS ' =
Tm
d x, A=
reactants
ki hi
products
ki hi . (3.22b)
32THE PHYSICS OF DEGRADATION
dN 'i
dx = (3.22c)
ki
n
Ni Ni n
S ' = R ln , N = N i . (3.24)
i N N i
1 1
dS ' = dQ . (3.25)
Tl Th
Rubbing one body against another induces friction forces between bod-
ies and simultaneously generates sliding wear. Wear, defined here as loss
of volume from material during rubbing, irreversibly reorders the rubbed
material, and as such, irreversible entropy must be produced. Friction
force F, and wear w, are manifestations of the same dissipative processes
that occur at the interface between rubbing bodies. Historically, friction
and wear were often treated as separate phenomena, but thermodynam-
ics and the DEG theorem admit a unified approach (Bryant, Khonsari
and Ling 2008). Many sliding wear systems operate at a steady state,
wherein dE = dS = 0, which suggests use of Equation 3.13. With no
reactions or phase change of material (dNk 0) within the control vol-
ume, Equation 3.13 asserts dissipated work to be the prevalent entropy
producer. Since the direction of the friction force vector always opposes
the direction of the sliding velocity vector during rubbing, with v = dx/dt
inFigure 3.1(a), friction dissipates power Fv and via Equation 3.13
dS/dt = Fv/T.
Through the DEG theorem of Equation 3.8, the rate of degradation
by wear is
dw dS ' Fv m Nv
=B =B =B , (3.27a)
dt dt T T
(a)
Slider
Tribo
control
y volume
dQ=TdeS
dx
x F
dt
Counter surface
(b)
1.0
0.8
Normalized wear
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Normalized entropy flow
mH m
k=B . (3.27b)
T
dw dS ' F dx 2 B dE m
=B = B [ 2( d e)( d f )] = 2 B = . (3.27c)
dt dt T dt T dt
S M = S M ( M ) (3.28)
K
Mk
M
k
k
1 (3.29)
f
0.7
0.6
0.5
Sn
0.4 3
10 2 4 6 8 104 2 4 6 8 105 2 4 6 8 106 2 4 6 8 107
Nkhg"N"*e{engu"*nqi++ S = 0.5S
n u
(in ksi, Sn 0.25 Bhn;
in MPa, Sn 1.73 Bhn)
states that the sum of the fractions Mk/ M kf of the life spent under each
load condition must not exceed unity. From this, the remaining number of
cycles under different conditions can be estimated.
Amiri and Khonsari (2012), Naderi and Khonsari (2010), and
Amiri, Naderi, and Khonsari (2011) have shown that fatigue of some
metals, composites (Naderi and Khonsari 2012), and polymers obey the
thermodynamic degradation paradigm and the DEG theorem. Collec-
tively, these references experimentally correlate the cumulative effects of
fatigue damage with the entropy produced in a fatiguing member. Amiri,
Naderi, and Khonsari (2011) formulated an expression for the irre-
versible entropy S generated during progression of fatigue, and mea-
sured entropy flow Se under an approximately stationary process dS = 0.
Through Equation 3.10, dSe = dS, which allows entropy flow to esti-
mate entropy generation. Naderi, Amiri, and Khonsari (2010) determined
entropy flow via infrared temperature measurements on a cantilevered
member undergoing reverse bending, and used finite element calculations
to estimate stress, the dissipated work of plastic deformations, the heat
released from these dissipations, and the entropy flow associated with the
heat flow over the fatigued member. The temperatures estimated by finite
element were consistent with the infrared measured temperatures.
During the first hundred fatigue cycles or so, temperatures increased
due to heat of plastic work dissipation (Naderi, Amiri, and Khonsari 2010;
Amiri, Naderi, and Khonsari 2011), see Figure 3.2. During the second
phase of thousands of cycles, temperatures stabilized at near constant
levels set by equilibrium between heat transfer and the dissipated work.
Finally, just before the end of the components fatigue life, temperatures
abruptly increased, and Naderi, Amiri, and Khonsari (2010) proposed
temperature to predict onset of catastrophic failure. Similar to Figure 3.1,
Naderi, Amiri, and Khonsari (2010), Amiri, Naderi, and Khonsari (2011),
and Naderi and Khonsari (2011) graphed normalized cycles M/Mf versus
normalized entropy Se/Sf, see Figure 3.3, where Mf and Sf are cycles and
entropy at specimen rupture. Collectively, the articles mentioned varied
fatigue conditionsgeometry, displacement amplitude, frequency, and
kinematicsand load typereversed bending and reversed torsion. All
cases rendered the linear relation
Se M
, (3.30)
Sf M f
K
S
k
k
S f . (3.31a)
dS ' S . (3.31b)
f
S M = S M ( S '). (3.32a)
40THE PHYSICS OF DEGRADATION
dS '
S M S0 = B dt = BdS '. (3.32b)
dt
U 0 U M f 1
U = U0 + ln(1 M M f ) (3.33b)
ln( M f )
1
In a log-log space for M << Me, Equation (3.28) will plot roughly as a straight line
of slope -a, similar to an integrator in a Bode magnitude plot. For M >>Me, the slope
will vanish due to the zero like term in the numerator, forming an S-N knee.
THERMODYNAMICS OF AGEING AND DEGRADATION 41
Electrochemical conversion
Ohmic phenomena
G0 G0 Eact Eact Vbat
0 E
G 1 TF 1 0 1
J J It I I I
Eact
Idt
act
It
Relec
G'
Ract Cdc
J
Electrochemical phenomena
G' Gstorage
0 1 Cstorage
Jdiff Jdiff
Energy storage
G'
Jcdiff
G
Jdiff
Cdiff Rdiff
Diffusion phenomena
internal cell impedance Z (ohms) (see Broussely et al. 2005), where the
internal cell impedance Z pertains to a Thevenin equivalent circuit of the
battery. Ageing reduces capacity C and increases impedance Z. Cycles
of charging and discharging age a battery. Battery cycle life is rated as
the number of complete chargedischarge cycles a battery can undergo
before (1) capacity C falls below 80 percent of initial rated capacity, and/
or (2) the internal resistance Z increases 1.3 to 2 times its initial value.
Battery life, typically 500 to 1,200 charge discharge cycles, depends
on many factors (Broussely et al. 2005 and Vetter et al. 2005). Most prev-
alent are:
The effects on battery life stated in the list are summarized in Table 3.1.
A battery operational model in bond graph form from Mnard,
Fonts, and Astier (2010) models the dynamic electrochemical phenom-
ena in a Li-ion battery. The bond graph of Mnard, Fonts, and Astier
(2010) was copied and is presented as Figure 3.3. A bond graph maps
42THE PHYSICS OF DEGRADATION
where and how power flows in a physical system, and where energy is
stored and d issipated. The half arrows in Figure 3.3 indicate the direc-
tion of positive power flow in a Li-ion battery. In a bond graph, poten-
tial energy is stored in capacitance elements C, kinetic energy is stored
in inertance elements I, and power is dissipated in resistance elements
R. From a completed bond graph, the differential equations that govern
the physics and dynamics of a system can be extracted (Brown 2006;
Karnopp, Margolis, and Rosenberg 2000). The bond graph of Figure 3.3
labels where in the battery system are the energy storage, diffusion
phenomena, electrochemical conversion, electrochemical phenom-
ena, and ohmic phenomena. On the far right of Figure 3.3, battery
terminal voltage Vbat and current I appear across the terminals of a phys-
ical battery. Chemical capacitance Cstorage in energy storage stores the
battery energy via electrochemical charge separation involving Li+ ions
and electrons. Gibbs free energy -DGstorage (J) and molar flow of lith-
ium ions J (mol sec1) appear as effort and flow on multiple bonds, indi-
cating the chemical thermodynamics embedded in this bond graph. The
THERMODYNAMICS OF AGEING AND DEGRADATION 43
minus sign on DGstorage refers to energy leaving the main storage Cstorage.
In diffusion phenomena, capacitance Cdiff and resistance Rdiff together
set a time constant Cdiff Rdiff, which controls the slow dynamics of Li+ ion
diffusion in the electrolyte, which transports charge through the electro-
lyte. Transformer TF:nF in electrochemical conversion has the modu-
lus of the Faraday constant F (9.649 104 C mol1) and the number of
moles of electrons n exchanged for each mole of lithium ions involved in
the electrochemical reaction at the electrodes. This transformer element
converts electrochemical power to electrical power. Effort source -G0
establishes a reference thermodynamic potential within the bond graph.
The power needed to activate electrochemical reactions at the electrode
electrolyte interfaces imposes resistance Ract. This power is dissipated.
Capacitance Cdc is due to a layer of charge (electrons and lithium ions Li+)
that forms about the electrodeelectrolyte interface. Low mobility of Li+
ions through the electrolyte relative to electron flows in the battery causes
ohmic resistance Relec.
In electric circuits, electric resistances dissipate power V I, where V
and I are the voltage drop across and current through the physical resistor.
The electric resistance Relec in Figure 3.3 has voltage drop helec and current
I. Similar to an electrical resistance, the power dissipated by other resis-
tances is the product of the equivalent voltages and currents, shown as
labels on the half arrow bonds. In the bond graph of Figure 3.3, resistances
Relec, Rdiff, and Ract dissipate powers helec I, -DGdiff Jdiff , and hact It, respec-
tively. The entropy generated dS is the dissipated power divided by a tem-
perature associated with the dissipative process. Thus from the resistance
elements and considerations associated with Equation 3.7,
dw Gdiff J diff h I h I
= Bdiff + Bact act t + Belec elec , (3.34b)
dt Tdiff Tact Telec
RTdiff RTact 2
Gdiff J diff = It , hact I t = It , helec I = Relec I 2 (3.35a)
a FI lim a FI o
giving
dw R R 2 R I2
= Bdiff I t + Bact I t + Belec elec . (3.35b)
dt a FI lim a FI o Telec
Here R is the molar gas constant, = 1/2, and Ilim and Io are diffusion cur-
rents dependent on the number of lithium ions. Cancellation of Tdiff and Tact
between numerators of Equation 3.35a and denominators of Equation 3.34b
rendered the first two terms of Equation 3.35b explicitly independent of
temperature. These terms, which pertain to electrolyte diffusion and elec-
trodeelectrolyte interface activation, should increase with temperature,
for the model to be consistent with row five of Table 3.1 and item (c) of
the list. This suggests that the coefficients Bdiff and Bact should monotoni-
cally increase with temperature, in a manner consistent with the observa-
tions. Note that this is consistent with remarks that followed Equation 3.5.
The last term of Equation 8b has Telec in the denominator, suggesting more
influence from this term at lower temperature, consistent with item list (d)
and the last row of Table 3.1.
2
With each increment of energy dissipated during each chargedischarge cycle an
increment of entropy must be produced, via the second law of thermodynamics. As
cycles accumulate, the entropy produced accumulates, and list item (a) suggests
battery life diminishes with increased entropy accumulation.
THERMODYNAMICS OF AGEING AND DEGRADATION 45
3.8CONCLUSION
REFERENCES
Batteries. Journal of Power Sources 146, no. 12, pp. 9096. doi: http://
dx.doi.org/10.1016/j.jpowsour.2005.03.172
Brown, F.T. August 2006.Engineering System Dynamics, A Unified Graph-
Centered Approach. 2nd ed. CRC Press.
Bryant, M.D. 2010. Unification of Friction and Wear. In Recent Develop-
ments in Wear Prevention, Friction and Lubrication, ed. G.K. Nikas, pp.
15996. Austin, TX: Transworld Research Network (Research Signpost).
ISBN 978-81-308-0377-7.
Bryant, M.D. and M.M. Khonsari. 2008. Application of Degradation-Entropy
Generation Theorem to Dry Sliding Friction and Wear. Proceedings of STLE/
ASME International Joint Tribology Conference. Miami, FL. IJTC2008-
71079.
Bryant, M.D., M.M. Khonsari, and F.F. Ling. 2008. On the Thermodynamics
of Degradation. Proceedings of the Royal Society London Series A 464, no.
2096, pp. 200114. doi: http://dx.doi.org/10.1098/rspa.2007.0371
Dai, Z., S. Yang, and Q. Xue. 2000. Thermodynamic Model of Fretting Wear.
Journal of Nanjing University of Aeronautics & Astronautics (China) 32, no.
2, pp. 12531.
Duyi, Y. and W. Zhenlin. 2001. A New Approach to Low Cycle Fatigue Damage
Based on Exhaustion of Static Toughness and Dissipation of Cyclic Plastic
Strain Energy during Fatigue. International Journal of Fatigue 23, no. 8, pp.
67987. doi: http://dx.doi.org/10.1016/s0142-1123(01)00027-5
Electropaedia. 2005. Battery and Energy Technologies: Battery Life (and Death).
http://www.mpoweruk.com/life.htm
Erdinc, O., B. Vural, and M. Uzunoglu. June 911, 2009. A Dynamic Lithium-Ion
Battery Model Considering the Effects of Temperature and Capacity Fading.
2009 International Conference on Clean Electrical Power pp. 3836. doi:
http://dx.doi.org/10.1109/ICCEP.2009.5212025
Fouvry, S. and P. Kapsa. 2001. An Energy Description of Hard Coating Wear
Mechanisms. Surface and Coating Technology 138, no. 23, pp. 1418. doi:
http://dx.doi.org/10.1016/s0257-8972(00)01161-0
Fouvry, S., P. Kapsa, H. Zahouani, and L. Vincent. 1997. Wear analysis in fret-
ting of hard coatings through a dissipated energy concept. Wear 203204,
pp. 393403. doi: http://dx.doi.org/10.1016/s0043-1648(96)07436-4
Fouvry, S., T. Liskiewicz, P. Kapsa, S. Hannel, and E. Sauger. 2003. An Energy
Description of Wear Mechanisms and Its Applications to Oscillating Sliding
Contacts. Wear 255, no. 16, pp. 28798. doi: http://dx.doi.org/10.1016/
s0043-1648(03)00117-0
Frederick, D. and T.S. Chang. 1965. Continuum Mechanics. Boston, MA: Allyn
and Bacon.
Huq, M.Z. and J.-P. Celis. 2002. Expressing Wear Rate in Sliding Contacts
Based on Dissipated Energy. Wear 252, no. 56, pp. 37583. doi: http://dx.
doi.org/10.1016/s0043-1648(01)00867-5
THERMODYNAMICS OF AGEING AND DEGRADATION 47
Vetter, J., P. Novak, M.R. Wagner, C. Veit, K.C. Moller, J.O. Besenhard, M. Win-
ter, M. Wohlfahrt-Mehrens, C. Vogler, and A. Hammouche. 2005. Ageing
Mechanisms in Lithium-ion Batteries. Journal of Power Sources 147, no. 1,
pp. 26981. doi: http://dx.doi.org/10.1016/j.jpowsour.2005.01.006
Whaley, P.W. November 915, 2012. Critical Entropy Threshold: An Irreversible
Thermodynamic Theory of Fatigue. Proceedings of the ASME 2012 Inter-
national Mechanical Engineering Congress and Exposition IMECE2012.
Houston, Texas.
CHAPTER 4
Thermodynamic Damage
within Physics of
Degradation
Alec Feinberg
SECTION 1: E
QUILIBRIUM THERMODYNAMIC
DAMAGE ASSESSMENT
4.1INTRODUCTION
DOI: 10.5643/9781606504673/ch4
50THE PHYSICS OF DEGRADATION
will cause dislocations to move into the semiconductor material; iron alloy
steel beams will start to corrode as oxygen atoms from the atmospheric
environment diffuse into the steel. In all of these cases, the spontaneous
processes creating disorder are irreversible. For example, the air is not
expected to go back into the bicycle tire; the semiconductor will not spon-
taneously purify; and the steel beam will only build up more and more
rust. The original order created in a manufactured product diminishes in a
random manner, and becomes measurable in our macroscopic world.
Associated with the increase in total disorder or entropy is a loss of
ability to do useful work. The total energy has not been lost but degraded.
The total energy of the system plus the environment is conserved during
the process when total thermodynamic equilibrium is approached. The
entropy of the ageing process is associated with that portion of matter that
has become disorganized and is affecting our devices ability to do use-
ful work. For the bicycle tire example, prior to ageing, the system energy
was in a highly organized state. After ageing, the energy of the gas mole-
cules (which were inside the bicycle tire) is now randomly distributed in
the environment. These molecules cannot easily perform organized work;
the steel beam, when corroded into rust, has lost its strength. These typical
second-law examples describe the irreversible processes that cause ageing.
More precisely, if the entropy has not increased, then as stated earlier,
the device/system has not aged. Sometimes, it will be helpful to separately
talk about entropy in two categories: (a) entropy damage as compared
with (b) general entropy change which does not cause device damage.
For example, the bicycle tire that has degraded due to energy loss did not
experience damage and can be reused. However, the corrosion of the steel
beam is permanent damage. In some cases it will be obvious, in other
cases we may need to keep tabs on entropy damage. In most cases, we will
mainly be looking at entropy change due to device ageing as compared
with absolute values of entropy, since entropy change is easier to measure.
Entropy in general is not an easy term to understand, it is like energy, the
more we learn how to measure it, the easier it becomes to understand.
The system is some sort of device set apart for study. From an engi-
neering point of view, of concern is the possible ageing that can
occur to it.
The environment is the neighboring matter, that interacts with the
system in such a way that it drives it toward its thermodynamic
equilibrium ageing state.
the ageing transistor. Improve the designs reliability so that less heat is
generated, or we can lower the environmental stress such as decrease the
power applied to the transistor. In the limit of reducing the stress to zero,
we approach a reversible process. Therefore, reversible processes must be
quasi-static. However, this does not imply that all quasi-static processes
are reversible. In addition, the system may be repairable to its original
state from a reliability point of view.
A quasi-static process ensures that the system will go through a gen-
tle sequence of states such that a number of important thermodynamic
parameters are well-defined functions of time, if infinitesimally close to
equilibrium (so the system remains in quasi-static equilibrium), the pro-
cess is typically reversible.
A repairable system is in a sense repairable-reversible or less irre-
versible from an ageing point of view. However, we cannot change the
fact that the entropy of the universe has permanently increased from the
original failure and that a new manufactured part had to be made for the
replaceable part. Such entropy increase has likely caused damage to the
environment that we live in.
e nvironment and partly due to the heat DQ that flows from the environ-
ment to the system
DU = DQ + DW (4.1)
In the case where heat and work are added to the system, then either
one or both can cause damage. If we could track this, we would find that
the portion of related damage is causing loss in the systems free energy
(which is discussed in Section 4.7).
If heat flows from the system to the environment, then our sign con-
vention is that DQ < 0. Similarly, if the work is done by the system on the
environment then our sign convention is that DW < 0. That is, adding DQ
or DW to the system is positive, increasing the internal energy. (Note that
the first law does not prohibit a degraded system from spontaneous repair,
which is a consequence of the second law.)
During the quasi-static process, the work done on the system by the
environment has the form
dW = Y dX
a
a a (4.2)
Prior to ageing, our device has a certain portion of its energy that is avail-
able to do useful work. This is called the thermodynamic free energy. The
thermodynamic free energy is the internal energy input of a system minus
the energy that cannot be used to create work. This unusable energy is the
entropy (S) multiplied by the temperature (T) of the system. There are two
thermodynamic free energies widely used. They are Helmoltz free energy
(F), which is the capacity to do mainly mechanical (useful) work, while
the Gibbs energy is primarily for non-mechanical work in Chemistry. The
systems free (or available) energy is in practice less than the systems
energy U; that is, if T denotes the temperature of the environment and if
S denotes the system entropy, then Helmholtz free energy is F = U TS,
which obeys F < U. The free energy is then the internal energy of a system
minus the amount of unusable (or useless) energy that cannot be used to
perform work. The unusable energy is given by the entropy of a system
multiplied by the temperature of the system. If the systems initial free
energy is denoted by Fi (before ageing) and the final free energy is denoted
by Ff (after ageing), then Ff < Fi. The system is in thermal equilibrium
with the environment when the free energy is minimized. We can also
phrase the second law for ageing in terms of the free energy
The second law of thermodynamics in free energy in terms of device
degradation: The spontaneous irreversible damage process that takes
place over time in a system decreases the free energy of the system toward
a minimum value. This spontaneous process reduces the ability of the sys-
tem to perform useful work on the environment, which results in ageing
of the system.
For an environment at a fixed temperature, the isothermal change
in the systems free energy is equal to the work done by the system on the
environment;
dQ
dS (4.6)
T
Here, the equality holds for reversible processes, and the inequality
for irreversible ones (as defined in the next section). Note that because the
entropy measurement depends on our observation of heat flow, which is
a function of the heat dissipation process (often termed path dependent in
thermodynamics), we again use the symbol dQ instead of dQ. The inequal-
ity in the above equation reminds us that the entropy change of a closed
system during an irreversible process is always greater than the entropy
transfer. That is, some entropy is generated during an irreversible process
(Sgen), and this generation is due to the presence of the irreversibilities. So,
instead of the inequality, we can write this as
2 dQ
S Sys = S 2 S1 = + SGen (4.7)
1 T
dU = dQ + dW = dQ + Ya dX a (4.8)
a
THERMODYNAMIC DAMAGE 59
U = U ( S , X 1 , , X n ), (4.9)
dU = TdS + Ya dX a (4.10)
a
S = S (U , X 1 , , X n ), (4.11)
1 Y
dS = dU a dX a (4.12)
T a T
At this point, the system has lost all its useful work, maximum damage
has occurred, and further ageing no longer will occur.
Consider a system that can exchange energy with its environment. The
system can exchange energy with the environment subject to the constraint
of overall energy conservation. The total conserved energy is written
U tot = U + U en (4.14)
dU tot = dU + dU en = 0, (4.15)
dU dU en
dStot = + (4.16)
T Ten
Since dU = dUen
1 1
dStot = dU , (4.17)
T Ten
dStot 0 (4.18)
1 1
dStot = dU = 0. (4.19)
T Ten
+Q
U tot = U + U en , (4.21)
qtot = q + qen (4.22)
obeying
dU tot = dU + dU en = 0, (4.23)
dqtot = dq + dqen = 0. (4.24)
Then we can write from the combined first and second law equations
dU Vdq dU en Edqen
dStot = + , (4.26)
T T Ten Ten
or equivalently
1 1 V E
dStot = dU dq , (4.27)
T Ten T Ten
dStot 0. (4.28)
dU = dW Po dV dQ (4.29)
dQ
d SGen = dS (4.30)
T
dW = dU Po dV To ( d SGen dS ) (4.31)
Integrating from the initial to final state we have for the actual work
Wr = ( U + Po V To S ) (4.35)
THERMODYNAMIC DAMAGE 63
Wr Wr
= 0, =0 (4.36)
U V V U
which gives
S S
1T = 0, Po To =0 (4.37)
U V V U
We have defined the maximum entropy and minimum free energy as equi-
librium states. We now use these definitions to formally define failure:
Catastrophic failure occurs when the systems free energy is as small
as possible and its entropy is as large as possible, and the system is in its
true final thermodynamic equilibrium state with the neighboring environ-
ment such that
dj dSTotal
= 0, and = 0 (4.40)
dt dt
64THE PHYSICS OF DEGRADATION
N
SGen = STotal = Si = S Sys + S Surroundings 0 (4.42)
i =1
66THE PHYSICS OF DEGRADATION
where the equality holds for reversible processes and the inequality
for irreversible ones. This is an important result for thermodynamic dam-
age. If we can keep tabs on DSTotal over time, we can determine if ageing
is occurring even in a complex system. To do this, we need a repeatable
method or process to make ageing measurements at different times. If we
find that the entropy has changed over time from a repeatable quasi-static
measurement process, then we are able to measure and track the ageing
that occurs between the systems initial, intermediate, and final states. We
can call this the entropy of an ageing process. (Note that, during system
ageing, we do not have to isolate the system. We only need to do this
during our measurement process.)
where the equal sign is for reversible process and the inequality is for
irreversible one. However, what portion of the entropy generated causes
degradation to the system and what portion does not? To clarify similar to
Equation 4.4,
There is really no easy way to tell unless we can associate the degra-
dation through a measurable quantity. Therefore, in thermodynamic dam-
age, we are forced to define Sdamage in some measurable way.
As well in thermodynamics, we typically do not measure absolute
values of entropy, only entropy change. Let us devise a nearly reversible
quasi-static measurement process f, and take an entropy measurement of
interest at time t1
been applied at times tinitial and tfinal, respectively. Determine if the resistor
has aged from the measurement process and what is the final value for the
resistor.
Before and after ageing at temperature T1, we establish a quasi-static
measurement process f to determine the entropy change of the resistor at
an initial time ti and final time tf. A simple method would be to thermally
insulate the resistor and pass a current through it at room temperature T2,
for a small time t and monitor the heat rise to temperature T3. The internal
energy of the resistor with work done on it is
This yields
and
dQ
S f (t1 + t ) S f (t1 ) = (4.51)
T
T3
mC p (T )dT T3
S f (t1 ) = = mC p avg Ln (4.52)
T2
T T2
T4
S f (t2 ) = mC p avg Ln (4.53)
T2
and
T4
S Damage (t2 , t1 ) = mC p avg Ln (4.55)
T3
T T
R(t f ) = 4 2 R(ti ) (4.57)
T3 T2
We note
R = r L/A (4.58)
N
STotal = Si
i =1 (4.59)
For a complex bank, if we are able to use the exact same measurement
process f as we did for Example 4.4 then the ageing ratio is still
T2
S (t2 , t1 ) = mC Avg Ln (4.61)
T1
where C is the specific heat (Cp = Cv = C) and for simplicity we are using
an average value for C.
Then for a system made up of similar type of parts in accordance with
Equation 4.61 the total damage entropy is
N
T2 N
STotal = Si = Ln mC i Avg i (4.62)
i =1 T1 i =1
1 ( x m) 2
f ( x) = exp (4.65)
2s 2 2 s 2
THERMODYNAMIC DAMAGE 73
Then when this function is inserted into the differential entropy equation,
the results is given by (http://en.wikipedia.org/wiki/Differential_entropy)
1
S(X ) = log(2 e s( x) 2 ) (4.66)
2
This is an important finding for system noise. We see that entropy for a
Gaussian noise system, the differential entropy is only a function of its vari-
ance 2 (it is independent from its mean ). Note, the results are also very
similar if the noise is log normally distributed. For a system that is becoming
noisier over time, the damage entropy can be measured in a number of ways
where the change in the entropy at two different times t2 and t1 is
1 s2
S Damage = St 2 ( X ) St1 ( X ) = S (t 2, t1) = log t22 (4.67)
2 s t1
Interestingly enough, noise engineers are quite used to measuring
noise with the variance statistic. One of the most common measurements
of noise is called the Allan variance. This is a very popular way to measure
noise and is in fact very similar to the Gaussian variance, given by
It is not surprising that noise and temperature are related. For example,
JohnsonNyquist noise power goes directly with temperature. Electri-
74THE PHYSICS OF DEGRADATION
cally, voltage and current are also functions of temperature. Simply put, as
temperature increases, so too will device noise. Therefore, these two state
variables are often related. However, each offers a unique way of charac-
terizing ageing at the system level.
So far we have two states variable for measuring ageing. Another very
obvious one that we should look at for consistency is the failure rate l. We
do not think of the failure rate as a thermodynamic state variable. In fact,
to any reliability engineer, it is second nature that the failure rate is indeed
a key degradation unit of measure. For consistency, we would like to put
it in the context of damage entropy to see what we find. For a complex
system, the most common distribution used comes from the exponential
probability density function
f ( x) = l exp ( lx ) (4.70)
where, for the exponential distribution E[x] = 1/l. For a system whose fail-
ure rate is constant over time (i.e., a property of the exponential distribu-
tion), the entropy is also constant. Thus, there is no change in the entropy
in such a case over time. However, if the environmental stress changes, we
can measure the entropy change. In this case, the entropy damage change
would be for this system under two different environments
DS = Ln(lE1/lE2) (4.72)
We have asserted that temperature, noise, and failure rate are some good
key system thermodynamic state variables. However, depending on how
THERMODYNAMIC DAMAGE 75
you define the system and the environment, they can also be state variables
for the environment. An IC engine is a combustion system that interacts
with the environment causing entropy damage to the environment. This
pollution damage and that of other systems, on a global scale, are unfor-
tunately measurable as meteorologist track global climate change. The
temperature (global warming) effect is the most widely tracked parameter.
However, it might also make sense from what we determined above to
look at environmental noise degradation. As our environment ages, we
might expect from our above finding that its variance will increase. Some
indications of environmental variance change are larger swings in wind,
rain, and temperature, causing more frequent and intense violent storms. It
might prove prudent to focus on such environmental noise issues and not
just track global temperature rise.
SECTION 2: N
ON-EQUILIBRIUM THERMODYNAMIC
DAMAGE ASSESSMENT
X2
dW = Ya dX a or W = Y dX (4.74)
X1
a
where we need to sum the work. If the path is well defined, we can integrate
over the path from X1 to X2. Each generalized displacement dX a is accom-
panied by a generalized conjugate force Ya. Recall that because work is a
function of how it is performed (often termed path dependent in thermody-
namics), we used the symbol W instead of dW indicating this. For a sim-
ple system, there is but one displacement X accompanied by one conjugate
force Y. Examples of basic conjugate work variables are given in Table 4.1.
Associated with work causing damage by the environment on the
device is a loss of free energy, which we presented in Equation 4.5 and
repeated here for convenience
= Area
X-Strain
clip back and forth. The cyclic thermodynamic work is converted into
heat and damage entropy in the system as dislocations are added at every
cycle, causing plastic deformation. This produces metal fatigue and
irreversible damage. This is illustrated in Figure 4.2. Ageing from such
fatigue is due to external forces, which eventually result in fracture of the
paper clip.
If we were trying to estimate the amount of damage done at each
cycle, it might be more accurate to write the damage in terms of the num-
ber of dislocations produced during each cycle. Then we could also esti-
mate the cumulative damage. However, this number is usually unknown.
Therefore, we take a thermodynamic approach.
For a simple system, the path in the ( X , Y ) plane corresponding to a
cycle as shown in Figure 4.3.
The work term is dWs = YdX . Recall that the plane can represent any
one of the conjugate work variable that can undergo cyclic work given in
Table 4.1. In the ( X , Y ) plane, the cycle is represented by a closed curve
C enclosing an area . The closed curve is parameterized (during the time
interval ti < t < t f of the cycle as a moving point ( X (t ), Y (t )) in the plane.
78THE PHYSICS OF DEGRADATION
For a cyclic process C in a simple system, the work done on the sys-
tem by the environment is given by the integral around a closed curve
tf
dX (t )
YdX = Y (t )
Work(cycle) =
C ti
dt
dt (4.76)
In Figure 4.3, the system does work on the environment if the point
in the plane transverses the curve in a clockwise fashion. The environment
does work on the system if the point in the plane transverses the curve in
a counterclockwise fashion.
m m
w di w Ti
Damage = i =1
and Effective Damage = i =1
(4.78)
Wd WT
The work damage ratio: This consists of the work performed to the
work needed to cause system failure. In system failure, we exhaust the
maximum amount of useful system work. To consistently find this damage
ratio, all work found must be taken over the same work path.
THERMODYNAMIC DAMAGE 79
The path is important as there are many ways to walk up a hill; there-
fore, we will always require the numerator and denominator to travel the
same path. For the damage expression, wdi is the cyclic work damage per-
formed over i cycles, and Wd is the total work damage performed to cause
system failure. This damage is created in the system by the environment
doing work on the system or vice versa. When we cannot identify in mod-
eling and/or in measuring the system damage part of the work performed,
then we can use the work performed. Then the ratio is not the true damage,
but it represents an effective amount of damage. In this case we write WT
as the total work performed to failure and wTi as the total cyclic work per-
formed over i cycles. When we are trying to define damage, we should try
and make the distinction. As a mechanical work example, damage might
be related to plastic strain; we usually can measure and/or model this.
At this point we will drop the subscript distinction, and just work with
the concept (but one should try and keep the distinction in mind when
working a specific problem). Then the cyclic damage can be written over
n cycles as
Yn dX n
Damage = n (4.79)
WFailure
When the Damage ratio value is 1, failure results for which we require
the total work summed along the path traveled in the numerator to equal
the full work to failure that occurred along the same path traveled in the
denominator.
In many cases, it is actually possible to sum the work for each cycle
in a consistent measurement process. For example, if Y is the stress
and X the plastic strain, we can use strain gages and monitor the stress
level and integrate out the area after each cycle using computer soft-
ware. Thus, an accurate estimate can be made of the damage occur-
ring. Inother cases, where such measurements are not possible, often
a technique called Miners rule (Miner 1945) is used. This often serves
as agood approximation of the effective damage and is discussed in
Example 4.6.
Similarly, if the work is noncyclic with only one kind of work
involved, and we are cautious to make sure the path traveled is identical
in the numerator as in the denominator than the thermodynamic damage is
If there are other types of work stress causing damage, then these
can be summed as long as the work path is the same. That is, we travel
up the hill on the same path but we may at times carry a different amount
of weight along the way. If we travel the path with one weight, we may
breakdown just as we reach the top of the hill, while the next time we go
up the path with more weight, we will breakdown sooner. The key is the
path traveled, and then we can consider different stresses as we travel this
path. Then we should be able to assess the second time up the hill if and
when we will breakdown relative to the breakdown value of 1. In the case
of other stresses we write
Work = Yi dX i (4.81)
i
tf
dX (t )
w = Y (t ) dt = f (Y , k , Ea ) t (4.82)
ti
dt
Here t is the time and f (Y, k, Ea) is some function of the environ-
mental stress, k and Ea are the specific constants related to the degrada-
tion mechanism, such as a specific power exponent and activation energy.
Then, according to Equation 4.79 or 4.80, if we have damage between two
different environmental stresses Y1 and Y2, and failure occurs for each at
times t1 and t2, respectively, then the damage value of 1 requires that
f (Y2 , k , Ea ) t 2 t2 f (Y1 , k , Ea )
Damage = = 1 or AFD (1, 2) = =
f (Y1 , k , Ea ) t 1 t 1 f (Y2 , k , Ea )
(4.83)
f (Y1 , k , Ea ) (4.84)
t i = t 1 AF (1, i ) and AFD (1, i ) =
f (Yi , k , Ea )
Equation 4.79 for cyclic damage may be used to derive Miners empirical
rule (Feinberg and Crow 2001; Feinberg and Widom 2000) that is com-
monly used for accumulated fatigue damage, and a number of other useful
expressions in damage assessment of devices and machines.
Consider a system undergoing fatigue as in our paper clip example,
which was bent back and forth a certain distance for three cycles. To find
the actual work we need to sum up the cyclic work area of each bend cycle
since both the stress s and strain e will change slightly as the paper clip
fatigues. If we use Stokes theorem, it demonstrates the work is related to
the cyclic area of each
3
w1 (cycle) = sde = d sde + d sde + d sde (4.86)
i =1 Areai Area 1 Area 2 Area 3
82THE PHYSICS OF DEGRADATION
Miner also empirically assumed that the work for n cycles of the same
cyclic size is all that is neededin our framework, this means
Using this assumption, then we average the three cyclic areas of the
same stress multiplied by three in our case
wi = 3 s i de (4.87)
Following this approximation, for any stress level we just count the
number of cycles so that the total thermodynamic work is
s1de + n2 s 2 de + n3 s3 de + ..
w( for k types of stress ) n1
K
= n1 w1 + n2 w2 + n3 w3 + = ni wi
i =1
(4.88)
K
ni wi
Effective Damage i =1 (4.89)
WFailure
or
n1 s 1de + n2 s 2 de + n3 s 3 de +
Effective Damage
WFailure (4.90)
n1 s 1de n2 s 2 de n3 s 3 de
= + + +
WFailure WFailure WFailure
THERMODYNAMIC DAMAGE 83
Using the approximation above, the total work to failure is the same
for each cyclic size type, along the same work path so that
So this yields
n1 s 1de n2 s 2 de n3 s 3 de
Effective Damage + + +
WFailure WFailure WFailure (4.92)
n1 s 1de n2 s 2 de n3 s 3 de
= + + +
N1W1 N 2W2 N 3W3
Giving
n1 s 1de n2 s 2 de n3 s 3 de
Effective Damage + + +
N1 s 1de N 2 s 2 de N 3 s 3 de (4.93)
n n n
= 1 + 2 + 3 +
N1 N 2 N 3
n1 n2 n3 k n
Effective Damage + + + = i (4.94)
N1 N 2 N 3 i =1 N
i
It can be difficult to always know the cycles to failure at any ith stress level
in Miners application. However, since we are on the same work path,
often, we can establish a time cycle acceleration factor between the stress
levels. If AFD and N1 are known, then the Miners rule can be simplified as
per Equation 4.84 for the ith stress using
Ni = AFD(1, i) N1 (4.95)
n1 n2 n3
Effective Damage + + +
N1 AFD (1, 2) N1 AFD (1, 3) N1
k ni (4.96)
=
i =1 AF (1, i ) N
D 1
k
w(cycle) = VdQ (4.97)
i =1 Areai
k ni (4.98)
Effective Damage
i =1 N
i
THERMODYNAMIC DAMAGE 85
104
P"?"6972"Gzr*202524"FqF'+."T?20;;9
Rd/Cekf
1000
P"E{engu"vq"hcknwtg
P"?"344"Gzr*2024"FqF'+."T?20;;;
100
Cnmcnkpg"OpQ4
10
0 20 40 60 80 100
FqF'"*S1S%"kp"'+
Cy life (MnO2)
Cy life (Pb-Acid)
Figure 4.4. Lead acid and alkaline MnO2 batteries fitted data (Fein-
berg and Widom 2000).
j
K BT f j DoD %
N j = No e = N jo e (4.99)
j = j j + n Q j (4.100)
where n is the activation damage voltage and Qj is that value for which the
battery manufacturer assumes the chemical cell to be fully discharged for
the jth battery type, f is a damage activation energy, and then we identify
jj
K BT
n Qj
N jo = N o e , and f j DoD% = (4.101)
K BT
86THE PHYSICS OF DEGRADATION
GM particles m Mi i (4.102)
Damage = = i
GN particles m N
i
i i
n (VI )i ti n rt n t
Damage ( corrosion ) = = i i = i (4.103)
i =1 (VI ) t i =1 r t i =1 t
i i i i i
where for the ith stress condition (for example, temperature stress i), ti is
the corrosion time, ti time to fail. Note we have introduced ri, the corro-
sion rate that goes as the corrosion current (see Equation 4.104). We find
the expression simplifies and we do not actually need to know the corro-
sion current or rate at any stress level, if we know the time to failure ti at
each ith level similar to the Miners rule. We can view this for general,
galvanic, or specific corrosion processes such as primary batteries.
Consider the case when the damage value is 1, failure occurs and
we can derive a useful expression for the acceleration factor in terms of
the stress. However, in this case, we will first need an expression for the
corrosion rate.
THERMODYNAMIC DAMAGE 87
r = ICorr B Eq / r A (4.104)
I1 t1 t I
Damage (corrosion ) = 1 = , then AF = 2 = 1
I 2 t2 t1 I 2
(4.106)
e 1 1
= exp (
R T1 T2
I1t1 t
Damage (corrosion ) = = AF (1, 2) 1 (4.107)
I 2 t2 t2
The reader should note that in general the rate constant K cancels out,
but depending on the way Equation 4.105 is written, and the test condi-
tions, that may not always be the case.
An alternative expression for primary batteries is based on the Peu-
kerts law (Linden and Reddy 2002).
Cp = I Yt (4.108)
88THE PHYSICS OF DEGRADATION
Y
t C p 2 I1
AF = 2 = (4.109)
t1 C p1 I 2
Ga Gc
i = i i = I exp I exp } (4.110)
corr a c oa RT oc RT
Here we have identified the barrier height in terms of the Gibbs free energy.
The thermodynamic work to change the free energy or equivalently cross
the barrier equates to DG = -W = -qE, where q = zH, H is the Faraday
constant, Eo is the open circuit cell potential, and z is the stoichiometric
number of electrons in the reaction, that is
M M Z + + ne (4.111)
where M is the metal forming M Z+ ions in solution. The anode and cathode
works is distributed (i.e., na = na, nc = n(1 - a) , often a = 0.5) so that the
anode work amount is, for example,
DGa = -a n H E (4.112)
THERMODYNAMIC DAMAGE 89
If2 na HE 1 1
AF = Exp (4.113)
I f1 R T2 T1
na HE (1 a )nHE
Coa K ao Exp o
= Cc K c Exp (4.114)
RT RT
Here we substitute into Equation 4.110 the anode and cathode ampli-
o o
tudes (Linden and Reddy 2002) Ioa = CoaK a nFA and Ioc = CocK c nFA. Ka
and Kc are temperature-dependent rate constants, C0 is the concentration
of the reducing agent at the anode, and CR is that of the oxidizing agent at
the cathode electrode surface.
Collecting terms yields the Nernst thermodynamic equilibrium
condition
RT CR C
E = Eo + ln or G = G o RT ln R (4.115)
nH Co Co
where
RT K f K of
o
Eo = ln o , Go = nHE o = RT ln o (4.116)
nH K b Kb
In aqueous corrosion (above) the currents are easier to measure than they
are in microelectronics occurring arbitrarily on a circuit board. In this
instance, the concept of adding the local percent relative humidity at or
near the surface has been found to aid in describing the potential for cor-
rosion to occur. The rate of corrosion and the rate of mass transport are
related then to the local percent relative humidity present at the surface,
which enhances the electrolyte at the surface for conducting the corrosion
currents. The corrosion current is not well defined but is proportional to
the rate kinetics and this local relative humidity (%RH) as
I Corr (% RH ) M K (T ) (4.117)
M
I Stress RH Stress K (T ) Stress
AFTH = = = AFH AFT (4.118)
IUse RHUse K (T )Use
E 1 1
AFT ,V = AFT AFV = Exp a (4.119)
K B TUse TStress
4.11THERMODYNAMIC DAMAGE IN
MECHANICAL SYSTEMS
Creep parameters include the strain (e) length change DL (e = DL/L) due
to an applied stress (s) at temperature (T ). In the elastic region, stress
causes a strain that is recoverable (i.e., reversible) so that s = Ye, where Y
is Youngs modulus. When stress increases such that yielding results, the
work is irreversible and the permanent plastic strain ep damage is termed
plastic deformation. The most popular empirical creep rate equation is
e p = Bo s M t p e Ea / KbT (4.120)
d eP
w = s d eP = s dt = Bs M +1 pt p 1dt = B s M +1t p (4.121)
dt
Table 4.3. Constants for stresstime creep law (Collins, Busby, and
Staab 2010)
Temperature B (in2/lb)N
Material o
C per day M
1030 Steel 400 48 1038 6.9
1040 Steel 400 16 1046 8.6
2Ni-0.8Cr-0.4Mo 454 10 10 20 3.0
Steel
12Cr Steel 454 10 1027 4.4
12Cr-3W-0.4Mn 550 15 10 16 1.9
Steel
92THE PHYSICS OF DEGRADATION
1 1 M +1 p
Ea / K B
T2 T1 s2 t2
Damage = e s t (4.122)
1 1
1
t Ea / K B ( T1 T2 ) s1
1 1 ( M +1) p
AFD = 2 = e s (4.123)
t1 2
p p
t ti
Damage = i = (4.124)
i ti i AFD (1, i ) t1
1
t Ea / K B ( T1 T2 ) s1
1 1 (M ) p
AF = 2 = e s (4.125)
t1 2
There are many different types of wear, these include abrasive, adhesive,
fretting, and fatigue wear. The most common wear model used for adhe-
sive and sometimes abrasive wear of the softer material between two slid-
ing surfaces is the Archards wear equation (Archard 1953; Archard and
Hirst 1956)
kPl
V= (4.126)
H
dD kP 2 n kP 2 n t
w = F dx = F dt = dt = (4.127)
dt HA HA
At some point, we may consider that too much damage has occurred
and define this as a failure amount. Then we can assess the damage ratio
if we have knowledge of the critical wear depth for a particular velocity n1
and stress level P1 (on the softer surface) and corresponding critical failure
94THE PHYSICS OF DEGRADATION
time t1. Then the damage at any stress level relative to a known failure
amount occurring, say, at stress level 1 is
2
P An t
Damage = 2 1 2 2 (4.128)
P1 A2 n1 t1
(Note that this is not an effective damage, because of the way Equation 4.126
is written in terms of measurable damage). When the damage ratio is 1, we
obtain the wear time compression acceleration factor
2
t P A n
AFD (2,1) = 2 = 1 2 1 (4.129)
t1 P2 A1 n 2
t ti
Damage = i = (4.130)
i ti i AFD (1, i ) t1
Note, since l = vt, we can substitute t for l, and t for L, if we wish to write
this in terms of the sliding distance rather than the wear contact time.
Note on Wear Damage Acceleration Factor versus Reliability Acceler-
ation Factor
We see from Equation 4.126 that the reliability wear amount acceler-
ation factor can be found as
t P A n
AFD (2,1) = 2 = 1 2 1 (4.131)
t1 P2 A1 n 2
e P = b o n p T j e Ea / K BT (4.132)
where we have substituted for the nonlinear stress sM = bDT j. The thermo-
dynamic work causing damage from the thermal cycle process is found
from the stressstrain creep area if the stressstrain relation could be plot-
ted. The cyclic work is then found similar to Equation 4.121
de
sde = s
w=
dn
dn = bT j +1 pn p 1dn = b T k n p (4.133)
1 1 p k
Ea / K B
T2 T1 n2 T2
Damage = e N T (4.134)
1 1
1 1 K
N Q / KB
T1 T2 T1
AFD = 2 = e T (4.135)
N
1 2
where k/p = K and Q = Ea/p. This ratio is the temperature cycle Coffin
Manson acceleration factors (Coffin 1974; Manson 1953; Manson 1966)
when the activation energy Q is small and the Arrhenius effect neglected.
For example, in solder joint testing, K = 2 for lead-free solder and about
2.5 for lead solder. The activation energy is about 0.123 that is typically
used. For example, if the used condition is stress level 1 cycle between
20 and 60 (DT = 40C, Tmax = 60C), while the test stress condition is stress
level 2 cycled between 20C and 100oC (DT = 120C, Tmax = 100C), then
Arrhenius AF = 1.58, while the CoffinManson AF = 9 with an overall AF
96THE PHYSICS OF DEGRADATION
of 14.2. In the case where we have 1 cycle per day in use condition, we see
that 10 years of use condition is about 260 test cycles.
The relation Equation 4.133 is also similar to the NorrisLanzberg
(Norris and Lanfzberg 1969) thermal cycle model, which also includes a
thermal cycle frequency effect that is used mainly when parts do not fully
reach extreme of each temperature.
Lastly, it should be noted that if we have a number of different stresses
applied, the general damage ratio can be written by accumulating the ther-
modynamic damage along the same work path as
p p
n ni
Damage = i = (4.136)
i Ni i AFD (1, i ) N1
In a similar manner to the above argument for thermal cycle, we can find
the equivalent mechanical vibration cyclic fatigue damage. In a vibration
environment, a vibration level depends on the type of exposure. In test-
ing, typically two types of environments are used, sinusoidal and random
vibration profiles. In sinusoidal vibration, the stress level is specified in
Gs, where G is a unitless quantity equal to the sinusoidal acceleration
A divided by the gravitational constant g. In random vibration, a similar
quantity is used, termed Grms (defined below). Consider first the plastic
strain (e) caused by a sinusoidal vibration level G stress (s) in the material.
The strain can be written similar to Equation 4.132
e = bo n pG j (4.137)
de
sde = s
w=
dn
dn = b G j +1n p (4.138)
THERMODYNAMIC DAMAGE 97
p b
n G
Damage = 2 2 (4.139)
N1 G1
b
N G
AFD = 1 = 2 (4.140)
N 2 G1
b b
N T G Grms 2
AFD = 2 = n1 = 2 (4.141)
N1 Tn 2 G1 Sinusoidal Grns1
Random
b
G1
N 2 = N1 = C1Gb k = C S b (4.142)
G
2
N = C S b (4.143)
1
f2 2
Grms = PSD( f )df (4.144)
f1
(4.145)
Grms W psd f
b/2
T W psd 1 N2
AFW = 2 = = (4.146)
T1 W psd 2 N1
We now ask: how does the free energy change as degradation work occurs?
Sometimes, the system path to the free-energy minimum is smooth and
downhill all the way to the bottom. For other systems, the path may
descend to a relative minimum, but not an absolute minimum, something
resembling a roller coaster. The path goes downhill to what looks like the
bottom and faces a small uphill region. If that small hill could be scaled,
then the final drop to the true minimum would be just over the top of the
small hill. The small climb before the final descent to the true minimum is
called a free energy barrier. The system may stay for a long period of time
in the relative minimum, before the final decay to true equilibrium.
Often, the time spent in the neighborhood of the relative minimum
is the lifetime of a fabricated product, and the final descent to the true
THERMODYNAMIC DAMAGE 99
j B j j B + j (4.147)
where the subscript B refers to the barrier and Df is the barrier height.
dp j
= n exp (4.148)
dt K BT
da j (a)
= n exp (4.149)
dt K BT
a2
j (a ) = j (0) + ay1 + y2 + ... (4.150)
2
where y1 and y2 are given by
j (0) 2 j ( 0)
y1 = and y2 = (4.151)
a a 2
When a << 1, the first and second terms in the Maclaurin series yield
da ay1
= n (T ) exp (4.152)
dt K BT
j ( 0)
where n (T ) = n o exp .
K BT
P
a= A ln[1 + B t ] for a << 1 (4.153)
P
K BT n (T ) y1
A= and B = (4.154)
y1 K BT
(a)
5
4
Amplitude Y
2 Y = Ln(1+7 t)
0
0 2 4 6 8 10
Time
(b)
0
1
Amplitude Y
Y = Ln(1+7 t)
2
5
0 2 4 6 8 10
Time
Figure 4.5. Examples of Ln(1+B time) ageing law with (a) similar to
primary and secondary creep stages and (b) similar to primary battery
voltage loss.
102THE PHYSICS OF DEGRADATION
early time. This curve is representative of many ageing and kinetic pro-
cesses such as crystal frequency ageing (Warner, Fraser, and Stockbridge
1965), corrosion of thin films, chemisorption processes (Ho 2006), early
degradation of primary battery life (Linden 1980), activated creep (see
Example 4.14), activated wear (see Example 4.13), transistor key param-
eter ageing (see Sections 4.11.5, 4.11.6), and so forth. The significance of
parametric logarithmic-in-time ageing can further be put in perspective as
it can be tied to catastrophic lognormal failure rates. This is discussed in
the appendix (Feinberg and Crow 2001; Feinberg and Widom 2000).
V{rg"K
ogvcnu
V{rg"KK
Ygct"coqwpv
ogvcnu
V{rg"KKK
egtcokeu
Unkfkpi"fkuvcpeg"qt"nqcf
after that over time. Type III typically observed in lapping and polishing
for surface finishing of ceramics.
In activation wear, we consider surface mass trapped in a potential
well. We need to apply enough thermal energy for mass to escape from the
chemical bonds. When a normal force is applied to the surface due to con-
tact wear under a constant velocity, the normal force P results in material
removal. We will start by taking a naive kinetic friction approach
dM j (M )
n = m KEff (T ) P = m KEff P exp (4.156)
dt K BT
dM j (M )
= g exp (4.157)
dt K BT
mk P
In this case g = . We have to now depart from our nave approach
n
as kinetic friction, although insightful, is not helpful enough in accurately
depicting the amplitude in wear. To be consistent and specify the model a
bit further, we can use historical information such as the Archards wear
model in Equation 4.126. We have a few options, we write the amplitude
kP nr M o
g= = (4.158)
H t
That is, we can either make use of the Archards amplitude or write
the amplitude in experimental terms of an initial average mass removal M o,
and t, an average time constant for removal. This amplitude can be found
experimentally as we will describe. In addition, we have added r, the den-
sity of the softer material to 4.125 for consistent units with mass removal
and the other parameters are defined in 4.125. Expanding the free energy
as a function of the mass removal, we have
104THE PHYSICS OF DEGRADATION
dj
j ( M ) j ( 0) + M + ... = j (0) + m M + ... (4.159)
dM
dM j ( 0) + m M m M (4.160)
= g (T ) = g o exp = g (T ) exp
dt K BT K BT
and f is the activation energy for the wear process, which can be found
experimentally in testing at different temperatures, as explained below.
Rearranging terms, and solving for the mass as a function of time t and
integrating provides a logarithmic-in-time ageing TAT model in terms of
mass removal over time
K BT g (T ) m
A= and B = (4.162)
m K BT
have a probability to hop over a potential barrier and weaken the crystal
metal lattice. Its hopping rate of occurrence is
dN j(N )
= N o exp (4.163)
dt K BT
de j (e)
= e o s N exp (4.164)
dt K BT
Note the difference from Equation 4.120 on the right hand side regard-
ing the time dependence of this popular form for the creep rate. Also the
key difference in this approach is that we associated the creep rate process
with the activation free energy as a function of the strain itself. We will at
this point use the TAT model by expanding it in terms of the strain depen-
dence in a Maclaurin series
dj
j ( e ) j (0) + e + ... (4.165)
de
We identify that the change in the free energy is due to damage that
occurred from mechanical thermodynamic conjugate work (see Table 4.1)
d j dW
= = s (4.166)
de de
de j (e) se
= e o s N exp = e 0 (T ) s N exp (4.167)
dt K BT K BT
This is now in the form of the TAT model (see Equations 4.153 and
4.154) we find in integration
L
e= A ln[1 + B t ] for e << 1 (4.168)
L
106THE PHYSICS OF DEGRADATION
and
K BT e (T ) s N +1
A= and B = o (4.169)
s K BT
or
L K BT e (T ) N +1
e= ln[1 + o s t ] (4.170)
L s K BT
j ( 0)
where e o (T ) = e o exp .
K BT
Here we illustrate how we can extend the TAT model to transistor age-
ing (Feinberg et al. 2000). We are primarily concerned with key transistor
device parameters. In the bipolar case for the common-emitter configura-
tion, the key transistor parameter is beta ageing, showing it to be directly
proportional to the fractional change in the base-emitter leakage current. In
the Field-Effect Transistor (FET) case, the key transistor parameter con-
sidered is transconductance ageing that results from a change in the drain
source resistance and gate leakage current. Then the TAT model is used
to provide an ageing expression that accounts for the time degradation of
THERMODYNAMIC DAMAGE 107
these parameters found in life test. These expressions provide insight into
degradation that links ageing to junction temperaturedependent mecha-
nisms. The mechanisms for leakage can be thought of similar to a corrosion
process having corrosion current. We have all the similar components
an anode, cathode, a conducting path, and an effective type of electrolyte.
Some typical life test data on HBTs and Metal Semiconductor Field-Effect
Transistor (MESFETs) are illustrated.
b (t ) = b o b (t ) (4.171)
i i i
i d b d I ce I ce I ce i I ce I be
b (t ) = = = 2 I be = b o (4.172)
dt dt I be I be I be I ce I be
I (t ) I (t ) I (t ) I (t )
b (t ) b o ce be b o be = b o be
I ce I be I be I be
(4.173)
In the above equation, DIce has been set to zero as no change in this
parameter is usually observed experimentally. We added the absolute
value sign since beta overall will degrade (b(t) = bo-Db(t)) as the leakage
current to the base increases. Thus, the first result is that the change in b is
directly proportional to the fractional change in the base-emitter leakage
current.
108THE PHYSICS OF DEGRADATION
where tb = RbeCbe is the time constant for steady-state excess charges in the
base-emitter junction. As discussed above, the base-emitter junction pri-
marily contributes to ageing effects. Along with this bulk effect is parasitic
surface charging Qs and leakage. We can also treat these using a simple RC
charging model. In this view, the surface leakage can be expressed as
Qbe = Qs + Qk (4.176)
As the transistor ages, Qbe increases along with Ib. Some of the
increase in Qbe is caused by the increase in impurities and defects in the
base surface and bulk regions due to operating stress.
The impurities and defects cause an increase in electron scattering, and
an increase in the probability for trapping and charging and eventual recom-
bination in the base. The above features lead to an increased leakage current.
In the capacitive model shown in Figure 4.7, incremental changes are
dQ = C dV = C R dI = dI (4.177)
I be (t ) Qbe (t ) Vbe (t )
b (t ) b o (4.178)
I = b o Q = b o V
be be be
THERMODYNAMIC DAMAGE 109
VC
C RL
0
'"Ejcpig"kp"dgvc
10
15
20
25
0 100 200 300 400 500
Cikpi"vkog"*jqwtu+
b (t ) I be (t ) qbe (t )
= = = A Log (1 + B t ) (4.179)
bo I be qbe
4.12.5TAT MODEL FOR TRANSISTORS AND
DIELECTRIC LEAKAGE
flows through the dielectric. These models stem from the fact that the work
required to create defects that increase leakage current is thermally activated
and there is a reduction in the free energy barrier for defect generation due
to the electric field lowering the barrier for defect creation. For example, in
a bond breakage model, the E-field is related to the bonds breakage through
the dipole energy. The energy barrier for creating defects has the form
dj
j ( q ) j ( 0) + q + ... (4.180)
dq
We identify that the change in the free energy is due to damage that
occurs from electrical thermodynamic conjugate work Table 4.1
d j dW
= =V (4.181)
dq dq
(j o a E N )
j = CE K exp (4.182)
K BT
where E is the electric field, a and C are constants. For the PooleFrenkel
(Harrell and Frey 1999) model K = 1, N = 1/2, for the Schottky effect (Sze
1981) model K = 0, N = 1/2, for the Thermo-chemical model K = 0, N = 1,
for the tunneling models K = 2, E = 1 (Sze 1981).
We can start our leakage TAT model consistent with Equation 4.182
and use Equations 4.180 and 4.181 for the free energy with N = 1, and the
general constant K as
dq ( j (0) + qV ) qV
= qoV K exp = q0 (T )V K exp (4.183)
dt K BT K BT
This is in the form of the TAT model results (see Equations 4.153 and
4.154), where we find
qbe I be
b = = A ln[1 + B t ] for Dbloss<< 1 (4.184)
qbe I be
THERMODYNAMIC DAMAGE 111
500
300
'K1K
170C
200
0
0 100 200 300 400 500 600 700 800
Vkog"*kp"jqwtu+
Figure 4.9. Life test data of gate-source MESFET leakage current over
time fitted to the ln(1 + Bt) ageing model. Temperatures are ambient.
Junction rise is about 30C (Feinberg et al. 2000).
K BT q (T )V K +1
where A = and B = or
I beV K BT
K BT q (T ) K +1
b (t ) ln 1 + o V t (4.185)
VI be K BT
j ( 0)
where qo (T ) = qo exp .
K BT
g (t ) = g g (t ) (4.186)
m o m
112THE PHYSICS OF DEGRADATION
where go, the initial value, is taken in the linear portion of the transcon-
ductance curve, is
I DS
g =
o VGS Vo (4.187)
Here, we use the linear portion of the curve for simplicity. Similar
results will follow for other portions of the curve. The time-dependent
function Dgm(t) is found from its derivative as
. dg d
g (t ) = m =
I
DS =
I
i
DS
I
DS
.V GS
( )
m
dt
dt V
V V V 2
GS o GS o V V
GS
(4.188)
or
. .
. .
.
g m (t ) =
I
DS
I
DS
V GS
=g
I DS
V GS
V V I V V o I V V
GS o DS GS o DS GS o
(4.189)
(4.190)
R I GS
g (t ) = g DS
m o RDS I GS I
GS o
decreases as ageing occurs and more defects occur in the channel causing
increased scattering.
Here, we wish to point out that similar to b degradation (a mechanism
that we have modeled as dominated by leakage), MESFET gate leakage
data, as shown, commonly follows a logarithmic-in-time ageing form as
well. This time dependence is illustrated in Figure 4.9. The TAT model for
gate leakage similar to Equation 4.185 is
I GS (t ) K T V K +1
B ln 1 + I o (T ) t (4.191)
I GS VI GS K BT
j (0)
where I o (T ) = I o exp .
K BT
APPENDIX
A second TAT model can be obtained for both the initial ageing period and
end of life using both terms in the Maclaurin expansion in Equations 4.149
and 4.150 and performing the integration. The results obtained in Feinberg
and Crow (2001) and Feinberg and Widom (2000) are
n (T )
a = x + b erf 1 exp( K 2 ) b t + erf ( K )
b
1
where erf and erf are the error function and its inverse and
x = y1 / y 2, b = 2 K BT / y 2, K = x / b, and b = 2 / 2
P
R Erf-1(K1Time+K2)
P0
R2
P
= ALog(1+b Time)
P0
tCritical
tc
Vkog
Figure 4.10. Ageing over all time t with critical values tc prior to
catastrophic failure 3.
time. What is interesting in this model is that the rate of initial ageing is
mathematically connected to its rate of final catastrophic behavior in this
model. This suggests that if the initial ageing process is truly understood,
a catastrophic prognostic may be possible!
REFERENCES
Archard, J.F. 1953. Contact and Rubbing of Flat Surface. Journal of Applied
Physics 24, no. 8, pp. 98188. doi: http://dx.doi.org/10.1063/1.1721448
Archard, J.F., and W. Hirst. 1956. The Wear of Metals under Unlubricated Con-
ditions. Proceedings of the Royal Society A-236, pp. 397410. doi: http://
dx.doi.org/10.1016/0043-1648(57)90017-0
Bhushan, B. ed. 2001. Modern Tribology Handbook, K. Kato, and K. Adachi. CRC
Press.
Chiou, Y.C., K. Kato, and T. Kayaba. 1985. Effect of Normal Stiffness in Loading
System on Wear of Carbon SteelPart 1: Severe-mild Wear Transition. ASME,
Journal of Tribology 107, pp. 49195. doi: http://dx.doi.org/10.1115/1.3261114
Coffin, L.F. 1954. A Study of the Effects of Cyclic Thermal Stresses on a
Cuctile Metal. Transaction ASME 76, pp. 92350. doi: http://dx.doi.
org/10.1080/01495739208946130
Coffin, L.F. 1974. Fatigue at High TemperaturePrediction and Interpretation.
James Clayton Memorial Lecture. Proceedings of the Institution of Mechan-
ical Engineers (London) 188, pp. 10927. doi: http://dx.doi.org/10.1243/
pime_proc_1974_188_014_02
Collins, J.A., H. Busby, and G. Staab. 2010. Mechanical Design of Machine Ele-
ments and Machines. 2nd ed. New York: Wiley.
Feinberg, A. January 2015. Thermodynamic Damage Measurements of an Operat-
ing System. Reliability and Maintainability Symposium.
THERMODYNAMIC DAMAGE 115
Feinberg, A., and A. Widom. March 1996. Connecting Parametric Ageing to Cat-
astrophic Failure through Thermodynamics. IEEE Transaction on Reliability
45, no. 1, p. 28. doi: http://dx.doi.org/10.1109/24.488913
Feinberg, A., and A. Widom. January 2000. Thermodynamic Extensions of Min-
ers Rule to Chemical Cells. Reliability and Maintainability Symposium.
pp. 34144. doi: http://dx.doi.org/10.1109/rams.2000.816331
Feinberg, A., and D. Crow, eds. 2001. Design for Reliability. CRC Press.
Feinberg, A., P. Ersland, V. Kaper, and A. Widom. 2000. On Ageing of Key Tran-
sistor Device Parameters. Proceedings - Institute of Environmental Sciences
and Technology, p. 231.
Feinberg, A., and A. Widom. June 2000. On Thermodynamic Reliability Engi-
neering. IEEE Transaction on Reliability 49, no. 2, p. 136. doi: http://dx.doi.
org/10.1109/tr.2000.877330
Harrell, W.R., and J. Frey. 1999. Observation of PooleFrenkel Effect Satura-
tion in SiO2 and other Insulating Films. Thin Solid Films 352, pp. 195204.
doi: http://dx.doi.org/10.1016/s0040-6090(99)00344-2
Hazewinkel, M., ed. 2001. Differential Entropy. Encyclopedia of Mathematics.
Springer.
Hirst, W. 1957. Proceedings of the Conference on Lubrication and Wear. Institu-
tion of Mechanical Engineers, London, p. 674.
Ho, Y-S. 2006. Review of Second-order Models for Adsorption Systems. Journal
of Hazardous Materials B136, pp. 68189. doi: http://dx.doi.org/10.1002/
chin.200648222 http://en.wikipedia.org/wiki/Differential_entropy
Linden, D., ed. 1980. Handbook of Batteries and Fuel Cells. New York:
McGraw-Hill.
Linden, L. and T.B. Reddy. 2002. Handbook of Batteries. 3rd ed. New York:
McGraw-Hill.
Lubliner, J. 2008. The Physics of Plasticity. Plasticity Theory. (see Chapter 2),
Dover.
Manson, S.S. 1953. Behaviour of Materials under Conditions of Thermal Stress.
NACA-TN-2933 from NASA. Cleveland, OH: Lewis Research Center.
Manson, S.S. 1966. Thermal Stress and Low Cycle Fatigue. New York:
McGraw-Hill.
Miner, M.A. 1945. Cumulative Damage in Fatigue. Journal of Applied Mechan-
ics 12, pp. A15964.
Norris, K.C., and A.H. Landzberg. 1969. Reliability of Controlled Collapse
Interconnections. IBM Journal of Research and Development 13, no. 3,
pp. 26671. doi: http://dx.doi.org/10.1147/rd.133.0266
Peck, D.S. 1986. Comprehensive Model for Humidity Testing Correlation.
Proceedings of the IEEE International Reliability Physics Symposium 24,
pp. 4450. doi: 10.1109/IRPS.1986.362110 MIL-STD-810G. Method 514.6,
Annex, A. October 2008.
Sze, S.M. 1981. Physics of Semiconductor Devices. John Wiley and Sons.
Warner, A.W., D.B. Fraser, and C.D. Stockbridge. 1965. Fundamental Studies of
Ageing in Quartz Resonators. IEEE Transaction on Sonics and Ultrasonics
12, no. 2, pp. 5258. doi: http://dx.doi.org/10.1109/t-su.1965.29361
CHAPTER 5
Monitoring Degradation in
the Field
Xiandong Ma
5.1INTRODUCTION
DOI: 10.5643/9781606504673/ch5
118THE PHYSICS OF DEGRADATION
For a fundamental eddy current testing, two inductive coils are usually
sufficientwith one acting as an exciter and the other a detector. The
excitation coil with a number of turns is excited by an alternating current
(AC) source, which generates a changing magnetic field in its vicinity.
This time-varying magnetic field interacts with the test sample and then
induces eddy currents. The eddy currents in the sample in turn generate a
secondary magnetic field, which opposes the primary magnetic field. The
interaction between these two fields alters the distribution of the magnetic
flux, thus resulting in an apparent change of the coil voltage. By measur-
ing the coil voltage change, the properties such as the electrical conduc-
tivity and magnetic permeability of the material, structure of the metals,
flaws, or fatigue cracks in metallic samples can be determined by using
analytical and/or experimental methods.
With more inductive coils placed along the periphery of an object
space, it is possible to tomographically image the distribution of materials
inside a region of interest. The sensor array used can contain excitation
and detection coils, which can either be dedicated to a particular function
(exciter or detector) or operate in both modes. By energizing an excitation
coil with an AC signal, a measurement can be obtained from detection
coils. This kind of projection is continued until the last excitation coil is
excited and measurement is taken. The measured data are then manip-
ulated using appropriate mathematical inversion techniques to create an
image of the internal object distribution.
Eddy currents induced inside the test sample by the applied magnetic
field attenuate with depth below the test sample surface. This attenuation
is mainly governed by the test objects electrical conductivity , magnetic
permeability , and the applied frequency f for a given test geometry. The
standard depth of penetration can be used to characterize this diffusion
phenomenon, which, for plane geometry, is mathematically given by
d = 1/ p f ms . Apparently, high-frequency measurements give informa-
tion regarding the properties adjacent to the surface, whereas low-frequency
testing probes the subsurface deeper inside the test sample. Multi-frequency
eddy current testing has been therefore commonly employed.
One of the applications has been to the characterization of microstruc-
tural changes of cellular metal foams. Cellular metal foams are a relatively
new class of material and have increasingly been recognized due to their
exceptional combination of mechanical, thermal, acoustic, electrical, and
chemical properties. To identify the healthy condition of porous metals for
an industrial application, evaluation of the foam electrical properties is of
paramount interest, which can be achieved through nondestructive eddy
current measurements. The induced eddy currents flowing in the metal
foams are affected significantly by the foam properties and, consequently,
measurement of the impedance change, defined as the voltage change with
respect to the excitation current, on the detection coil(s) permits the metal
foams to be characterized.
Figure 5.1 shows solenoidal coils designed to examine the cylindri-
cally shaped samples (Ma, Peyton, and Zhao 2005a). It has been found that
the phase-frequency response of the normalized eddy current signal of the
sensor is relatively immune to the coil-to-sample spacing and fill-factor
variations, from which the equivalent conductivity of the foams are mea-
sured. As an example, Figure 5.2 shows the variations of electrical conduc-
tivity of the aluminum foams with both porosity and pore size. The results
show that higher porosity leads to lower equivalent electrical conductivity
due to the decreasing the metal volume ratio in the foams. The pore size
determines the amount of air trapped, the average wall thickness between
pores, and the degree of interconnectivity between pores, thus affecting
the equivalent conductivity of the foams as well. Consequentially, through
monitoring of the equivalent conductivity, the mean porosity and hence
health conditions of the foams can be estimated (Ma and Peyton 2006a).
The sensors and technique can also be used for measurement of magnetic
permeability of porous Fe samples (Ma, Peyton, and Zhao 2006b).
Decarburization is a well-known phenomenon that can change the
structure and content of steel. It occurs usually when steel is heated in an
environment where oxygen is present, leading to oxidation and loss of car-
Monitoring Degradation in the Field 121
Detection coil r1
h1 h2
Excitation coil r2
Sample
bon. As a result of decarburization, the steel can lose some of its strength
and ductility, and bring dramatic modifications to its mechanical properties,
like being vulnerable to breaking hence decreasing the fatigue lifetime of
the steel. Eddy current techniques have been demonstrated to be very sen-
sitive and well related to surface carbon content (Amiri and Kashefi 2009)
and surface modifications associated with the decarburizing (Mercier etal.
2006). The measurement of three different electromagnetic properties (eddy
current, tangential magnetic field, and incremental permeability) was used
to evaluate decarburization profiles in spring steel (Burzic, Zamberger, and
Kozeschnik 2010). The results, however, were highly influenced by the
calibration procedure. A multifrequency, multi-output electromagnetic sen-
sor was proposed with which hardness profile beneath the surface can be
determined by correlating an optimum frequency to each measured output
voltage (Kahrobaee and Kashefi 2011).
122THE PHYSICS OF DEGRADATION
F: 1000/2000, 72.8%
D: 710/1000, 72.7%
I: >2000, 83.4%
H: >2000, 74.2%
B: 425/710, 73.5%
80
Porosity (%)
75
C: 710/1000, 64.7%
E: 1000/2000, 63.8%
A: 425/710, 63.1%
G: >2000, 63.7%
70
65
0 1 2 3 4
Conductivity (MS/m)
Target
Active side
sensing flux
Active pickup coil
Common
excitation coil
Active side
Dummy side
Ferrite core
Dummy side
sensing flux
Figure 5.3. Magnetic head before encapsulation and the main elements
of the magnetic sensing head.
Figure 5.5. EMT images of steel flow profiles through the nozzle at
different time instants.
5.3INSULATION DEGRADATION
5.3.1 MULTIPLE PHYSICAL STRESSES
V(t)
Applied voltage
Time
0
Vi
Actual voltage
across void
Figure 5.6. Voltages and current pulses at a void or cavity in the dielectric
material.
(Arora and Mosch 2011). When the voltage V(t) is sufficiently low, no
discharges occur. As V(t) is gradually raised to the breakdown voltage
Vi, also known as discharge inception voltage, a discharge begins. The
discharge will be self-extinguished as the charge initially transferred
across the void builds on the void surfaces and reduces the electric field.
Further discharges occur when the decrease in field is counteracted by
a continuing rise in voltage until the peak of V(t), after which the field
will be dropping within the remaining period of this half cycle. A simi-
lar discharge sequence occurs during the negative half cycle. However,
due to the impact of the accumulated charge from the positive half cycle
and the applied voltage V(t), the electric field will be intensified in the
void during the negative half cycle, thus accelerating the breakdown that
occurs earlier on the applied wave. This discharge sequence happens con-
tinuously with the applied voltage V(t).
A three-capacitor model has been widely used to model a discharge
resulting from a gas-filled void inside solid or fluid dielectric materials,
as shown in Figure 5.7. Electric fields within the void are represented by
Cc and those starting at the void walls toward the dielectric materials in
series with the void form the equivalent capacitance Cb. All fields outside
the void, that is, in parallel with the void, are represented by the equiv-
alent capacitance Ca. Normally, Ca >> Cc >> Cb for realistic geometric
dimensions.
The charge accumulated at the measuring leads is referred to as the
apparent charge q transferred from the actual charge in a void. The apparent
130THE PHYSICS OF DEGRADATION
Vb
Cb
Dielectric Cb
in series Va Ca
Void Cc
with Cc Vc Cc
B
Dielectric Ca in parallel with Cc
where q = CbVc, Vc is the voltage drop across the void. It has been found
that there is a direct relationship between the apparent charge q and the
energy involved in the discharge, which causes the deterioration of the
dielectric, and a relationship between q and the size of the void. Therefore,
the apparent charge q is chosen as a measure for discharges, which can
readily be measured with electrical discharge detectors through measuring
Va at external leads of the test specimen. The apparent charge q is mea-
sured typically in picocoulomb pC. However, Ca and Cb need to be known
in order to calculate q.
5.4 PD MEASUREMENT
Thermography techniques
Given that PDs are generally hotter than their surrounding media,
it might reasonably be thought that the thermal imaging techniques
could be applied in their detection and measurement. Unfortu-
nately, most PDs are enclosed in some way, for example, within
solid insulation or within metal-clad enclosures. As the relatively
low temperatures of PDs and the high thermal impedances are
likely present between the discharges and the imaging technique,
the use of this type of testing is very limited. These techniques are
Monitoring Degradation in the Field 131
only applied in the cases where the discharges are external to asso-
ciated plant items, for example, on overhead lines/busbars or post/
string-type insulators. However, such techniques are carried out not
in any quantitative way.
Chemical detection
PD activity can cause chemical reactions in insulating media. In
power transformers, PDs manifest themselves in discharge-gener-
ated gases in the oil insulation and in degradation bi-products in
paper insulation.
Under the action of PDs, oil degrades through bond scission
to form characteristic gases dissolved in the oil. Typical gases pro-
duced are hydrogen, methane, ethane, ethylene, and acetylene. The
quantity and types of gases produced depend on the nature of PD,
its severity, and the associated temperature. By sampling oil in the
transformer and further analyzing the types of the fault-generated
gases and their corresponding constituents, dissolved gas analy-
sis (DGA) can be utilized to infer PD activity in the transformers
(Lelekakis et al. 2011). The DGA technique might be the simplest
and cheapest technique available for online testing of transformers
and has been used as the primary indicator of malfunctions leading
to overheating, arcing, and PD.
Several interpretation techniques have been developed and are
used in the interpretation of dissolved gases in oil. Although the
presence of PD in oil-filled plant item can be inferred from the abso-
lute levels of the measured different gases, PD is indicated primar-
ily through the ratios of these gases. Gas ratios have been utilized
to differentiate between fault types since 1970, typical example of
which is the Rogers ratios. Rogers (1978) used the order of gas
evolution to form the gas ratios of methane/hydrogen, ethane/meth-
ane, ethylene/ethane, and acetylene/ethylene. A diagnosis table was
created based on nearly 10,000 DGA results, together with exam-
ination of the failed units and the units with suspected faults. The
table went through several evolutions and was produced in two for-
mats; in the first, diagnosis was based on codes generated by ratios,
while, in the second, diagnosis was made based on the value of the
ratio. Different types of PD can be identified from the gas ratios.
However, the diagnosis obtained using gas-ratio methods was
often not comparable between techniques. Other additional graph-
ical techniques were, therefore, introduced to complement the
interpretation of dissolved gases in oil. Duval (1989) developed a
triangle based on the relative percentage of methane, ethylene, and
132THE PHYSICS OF DEGRADATION
integrity of the system begins to lose. As losses increase, the resistive cur-
rent becomes larger and an angle created between the capacitive current
and the resultant resistive current increases, as does the tan . By monitor-
ing the tan of the system, relative discharge activity can be inferred. Tra-
ditionally, the tan measurement employed a Schering Bridge type circuit
connected across the high-voltage terminals of the plant item, but latterly,
instruments have made direct measurements of the different current com-
ponents. The tan measurement is dependent on the capacitance and resis-
tance values in the equivalent circuit of test samples, the type of equivalent
circuit used, and the frequency of the external test voltage (Kuffel, Zaengl,
and Kuffel 2000). Any sudden change in tan would be considered indic-
ative of PD inception and the rate of change with increasing voltage would
be indicative of the relative severity of the PD. Because tan measures the
integrated quantities associated with the dielectric loss, this approach is
only capable of indicating the integrity of the insulating system degradation
of the specimens under test. It cannot distinguish between a few large dis-
charges that might be detrimental and a lot of small ones that might be quite
innocuous. Therefore, most commercial instruments have been developed
to incorporate a peak magnitude detector to deal with this problem.
One of the instruments is the partial discharge analyzer (PDA) devel-
oped by the Canadian Ontario Hydro. For rotating machines, a capacitance
coupler can be permanently connected to the machine phase terminals to
allow winding discharge measurements. In large hydraulic generators,
two coupling capacitors, each having a typical capacitance of 80 pF, are
required to be permanently installed at the ends of the ring bus of each
winding phase, respectively. The PDA instrument (Stone et al. 1990)
can take the signals from the pair of couplers and sort the PD pulses by
removing external noise by means of differential amplifier, thus making
the analysis of discharges in stator windings more reliable. The lengths of
the coaxial cables connecting the couplers at the ends in the same phase
to the terminals of PDA instrument are purposely designed to ensure that
the electrical distance from the machine terminal along the ring bus to
the instrument terminal is the same for the two couplers. The connection
scheme is shown in Figure 5.8, where two identical capacitive couplers C1
and C2 (80 pf) are permanently installed to capture signals. To ensure that
the test presents an equivalent electrical path to the inputs of the differ-
ential amplifier, the coaxial cables on both sides are designed with such a
specific relationship
b d
a+ = c+ (5.2)
0.65 0.65
Monitoring Degradation in the Field 135
Cable Differential
C1 amplifier
a
50
Ring bus
To measuring
c 50 instrument
Neutral
Cable
C2
d
enclosure of the plant and further to the atmosphere outside the switch-
board through a gap somewhere in the enclosure. A transient earth voltage
(TEV) is produced by the action of the wave connecting with the earthed
metalwork and can be detected by a capacitive probe (Brown 1996). This
method gives an indication of the PD severity with the magnitude of a
detected signal, which lies normally in the milli-volts to volts range. With
multiple capacitive probes used, it is possible to gain some knowledge
of location using the time of flight principle, as the probe nearest to the
source should detect the discharge first. The attractive advantage of this
type of testing is its nonintrusive nature, that is, no disruption or outage
is required to the plant under test. However, as with other types of PD
measurement, the measurement readings of using TEV principle suffer
readily from external electromagnetic noise. It is therefore essential that
a background survey should be completed prior to measurements so that
discharges from the plant or from external sources can be differentiated.
Antenna is another type of technique that could be used for on-line
monitoring of the PD severity of the plant (Kaneko et al. 2009). An
antenna measures the radiated radio interference generated by discharges
based on the principle that the frequency band of an antenna can be tuned
to a specific radio frequency bandwidth to best respond to the discharges.
As with the TEV method, antenna techniques are of nonintrusive nature.
However, PD signals, when measured with this type of testing, are also
highly susceptible to noise from environments in substation. Therefore,
an effective noise reduction technique is desirable when the antenna tech-
nique is applied. As a direct connection to high-voltage terminals is not
required, the PD measurements using the antennas can be made and car-
ried out without planned machine outage.
Current transformer (CT) has been used in particular for motors and,
to a lesser extent, for transformers as a first pass technique by engineers.
By connecting a clamp-on CT to the neutral strap of the plant item, the
output readings can be taken to an oscilloscope or a frequency spectrum
analyzer for further PD analysis. The use of the CT approach is extremely
cheap, simple, and safe, and no disconnections need to be made during
test. Unfortunately, this type of testing suffers from the inherent disad-
vantages mainly including the lack of an effective calibration method to
determine the magnitude of any present discharges, susceptibility to extra-
neous interference, for example, pulses from power electronics circuitry
and corona discharging from somewhere in the system, and no effective
phase information on the location of discharges with respect to the AC
voltage power cycle. In addition, there may not always be a neutral avail-
able on the plant item under test.
Monitoring Degradation in the Field 137
Rogowski coils have been used for the detection and measurement of
PD currents for decades (Hashmi, Lehtonen, and Nordman 2011; Zhang,
Xiao, and Li 2009). They work on Amperes law and are used to measure
currents via the magnetic field produced by these currents. An air-cored
coil is placed around the conductor in a toroidal fashion. The current flow-
ing through the conductor produces an alternating magnetic field around
the conductor, resulting in a voltage being induced in the coil. The rate of
change in this voltage is proportional to the rate of change in current. This
voltage is then integrated electronically to provide an output proportional
to the current. Rogowski coils are normally connected to the high-voltage
terminals of the plant items to measure the PD current output. In use, no
direct connection to the high-voltage terminals and no coupling capaci-
tor are required. However, this detection method does not work through
shields.
Another approach connecting the discharge transducer to the
high-voltage terminals of the plant item is the use of a capacitor divider
type assembly (Haddad and Warne 2007). Typically, this discharge trans-
ducer consists of a discharge-free high-voltage capacitor connected in
series with or in parallel with a low-voltage impedance circuit (RC or
RLC), which in turn is connected to an oscilloscope or similar data acqui-
sition instrument. The high voltage can be reduced to a safe level at the
low-voltage impedance (typically 1,000:1 ratio) by careful choice of com-
ponent values. Individual pulses from discharges can be displayed super-
imposed on the AC power cycle voltage.
Applied voltage
vi
qi:discharge
magnitude i :phase angle
Time (ms)
0 10 20
Figure 5.9. The PD pulse sequences and the relative parameters used to
characterize the recurrence of discharges.
20
10
0
0
326 360
300
Cr 653
rc 240
tg
pv 979 180 +
"ej 120 u"*
ct 1305
pf qy
ig 60 "yk
"*o 1631 0 cug
X+ Rj
as for gas-insulated switchgears (GIS) (Gao, Ding, and Liu 2011) and
power transformers (Beltle, Muller, and Tenbohlen 2012). However, the
phase-resolved PD methods are normally employed at field sites where
the phase information is available, which is not always the case, for exam-
ple, the PD patterns of free-moving conducting particles in GIS. A chaotic
analysis of PD was proposed in a study by Koo et al. (2010) to identify
the type of defects by means of PD patterns without employing the phase
information.
discharges occurring outside the test circuit are rejected. The other benefit
associated with this system is its capability of suppressing some internal
noise, such as static exciter transients, because this type of noise has a
relatively low-frequency range and its traveling time suffers little change
due to difference in signal path lengths.
Many researches have been reported on denoising of PD signals using
digital processing methods, especially for on-site PD measurements. Tra-
ditionally, the techniques used for digital signal processing are realized in
either the time or frequency domain. In the frequency domain, denoising
has been widely implemented by FFT coefficients to a certain threshold
for different frequency bandwidths (Ma, Zhou, and Kemp 2000; Sriram
et al. 2005). However, the FFT does not provide information on time,
such as the time at which changes in frequency occur. The windowed Fou-
rier transform, also known as the short-time Fourier transform (STFT), in
attempting to overcome this deficiency, can provide a two-dimensional
representation of a signal in the time-frequency domain through window-
ing the signal. However, resolution in both time and frequency remains
constant due to the same window being employed in STFT across the
entire frequency range. With regard to the PD signals, there always exist
nonperiodic and fast transient features, which tend to be ignored and can-
not be revealed efficiently and explicitly. For these reasons, the applica-
tions of Fourier transform method to denoising PD signals are limited.
Adaptive digital filters have been used in PD measurements directly in the
frequency domain for suppressing narrow-band sinusoidal interference
(Kopf and Feser 1995; Sher et al. 1995). However, white noise still con-
tinues to remain in the PD signals following digital filtering of the given
signals.
More recently, the wavelet-based denoising techniques have been
found to be more promising for detecting PD signals buried in noise. As
we know, the objective of denoising is to remove noise from the PD signals
as effectively as possible while preserving the signal features essential to
the application. The wavelet transforms can decompose/reconstruct a given
signal in varying scales and characterize the signal in both the time and
frequency domain simultaneously with variable resolution. Wavelet-based
denoising is accomplished by selecting a wavelet that can be scaled to rep-
resent effectively the signal features, performing a decomposition of the
signal using the selected wavelet, thresholding the wavelet coefficients
to eliminate those associated with noise while preserving those that are
required to describe the signal of interest, and then reconstructing the
denoised signal. For this reason, it is particularly suitable for the analy-
sis of transient, irregular, and non-periodic signals such as in the case of
142THE PHYSICS OF DEGRADATION
PD signals (Dey et al. 2010; Ma, Zhou, and Kemp 2002a; Ma, Zhou, and
Kemp 2002b). As an example, Figure 5.11(a) gives the wavelet denoising
results of a practical PD data set where PD signals at levels comparative
to and below noise can be fully extracted. Figure 5.11(b) shows the PD
pattern that is dominated by noise when raw data are directly used, whereas
Figure 5.11(c) shows the PD pattern based on the wavelet-based noise sup-
pression. The wavelet denoising technique has both eliminated noise and
detected more PD activity with low apparent discharge magnitude, and is
thus much more effective in characterizing the nature of the discharges.
However, denoised signals using wavelet methods can achieve good
results if the PD pulses are residing at different resolution levels with noise
and proper selection of mother wavelet, the number of decomposition lev-
els, and threshold value. It can be therefore envisaged that the method may
remove some useful information related to PD pulses if these conditions
are not fulfilled. Adaptive noise cancelation methods were used for pre-
venting degradation of the denoised PD signals. For example, the method
proposed in Kopf and Feser (1995) defines and weights the disturbing fre-
quency components involved in the measured PD signal and then remove
them. Although the proposed filters are self-configurable, they are nor-
mally realized by a high-order finite impulse response (FIR) filter and thus
are time consuming. An integration of artificial intelligence methods with
the wavelet denoising method has been proving a more robust approach
for PD detection and characterization. The applications include the use of
neural networks or neuro-fuzzy networks for pattern recognition to auto-
matically classify PD sources (Ma et al. 2013; Mazzetti et al. 2006).
5.5CONCLUSION
(a)
Voltage, mV
0
5
10
0 1 2 3 4 5 6 7 8 9
5
Voltage, mV
0
5
10
0 1 2 3 4 5 6 7 8 9
Time, ms
(b)
Number of discharges
300
200
100
0 360
0 270
4
App 180 ()
are 8 o ws
nt c ind
har 11 90 ew
g e (p
C) 15 P has
0
(c)
Number of discharges
30
20
10
0 360
0 270
4
App 180 ()
are 8 o ws
nt c ind
har 11 90 ew
ge (p
C) 15 Phas
0
Figure 5.11. (a) Raw data and denoised signal using wavelet, (b) PD
pattern of original data, and (c) PD pattern with wavelet-based noise
suppression.
144THE PHYSICS OF DEGRADATION
REFERENCES
Abu-Siada, A., S. Hmood, and S. Islam. 2013. New Fuzzy Logic Approach for
Consistent Interpretation of Dissolved Gas-in-oil Analysis. IEEE Transac-
tions on Dielectrics and Electrical Insulation 20, no. 6, pp. 234349. doi:
http://dx.doi.org/10.1109/tdei.2013.6678888
Amiri, M.S. and M. Kashefi. October 2009. Application of Eddy Current Non-
destructive Method for Determination of Surface Carbon Content in Carbu-
rized Steels. NDT & E International 42, no. 7, pp. 61821. doi: http://dx.doi.
org/10.1016/j.ndteint.2009.04.008
Andrews, T., R.N. Hampton, A. Smedberg, D. Wald, V. Waschk, and W. Weissenberg.
2006. The Role of Degassing in XLPE Power Cable Manufacture. IEEE Elec-
trical Insulation Magazine 22, no. 6, pp. 516. doi: http://dx.doi.org/10.1109/
mei.2006.253416
Arora, R. and W. Mosch. 2011. High Voltage and Electrical Insulation Engineer-
ing. IEEE Press.
Bai, L., G. Yun Tian, A. Simm, S. Tian, and Y. Cheng. March 2013. Fast Crack
Profile Reconstruction Using Pulsed Eddy Current Signals. NDT & E Inter-
national 54, pp. 3744. doi: http://dx.doi.org/10.1016/j.ndteint.2012.11.003
Batchelor, A.W., L. Nee Lam, and M. Chandrasekaran. 2011. Materials Degra-
dation and Its Control by Surface Engineering. 3rd ed. Singapore: Imperial
College Press.
Monitoring Degradation in the Field 145
Ma, X., C. Zhou, and I.J. Kemp. 2002. Automated Wavelet Selection and Thresh-
olding for PD Detection. IEEE Electrical Insulation Magazine 18, no. 2,
pp. 3745. doi: http://dx.doi.org/10.1109/57.995398
Ma, X., C. Zhou, and I.J. Kemp. May 1820, 2004. Novel Computer-based Pro-
cessing System for Partial Discharge Detection and Diagnosis. Proceedings
of the 21st IEEE Instrumentation and Measurement Technology Conference,
Vol. 1, pp. 6848. Como, Italy.
Ma, X., A.J. Peyton, and Y.Y. Zhao. July 2005. Measurement of the Electrical
Conductivity of Open-celled Aluminium Foam using Non-contact Eddy Cur-
rent Techniques. NDT & E International 38, no. 5, pp. 35967. doi: http://
dx.doi.org/10.1016/j.ndteint.2004.10.003.
Ma, X., A.J. Peyton, R. Binns, and S.R. Higson. April 2005. Electromagnetic
Techniques for Imaging the Cross-section Distribution of Molten Steel
Flow in the Continuous Casting Nozzle. IEEE Sensors Journal 5, no. 2,
pp. 22432. doi: http://dx.doi.org/10.1109/jsen.2004.842443
Ma, X. and A.J. Peyton. April 2006. Eddy Current Measurement of the Electrical
Conductivity and Porosity of Metal Foams. IEEE Transactions on Instru-
mentation and Measurement Technology 55, no. 2, pp. 5706. doi: http://dx.
doi.org/10.1109/tim.2006.873549
Ma, X., A.J. Peyton, and Y.Y. Zhao. 2006. Eddy Current Measurements of
Electrical Conductivity and Magnetic Permeability of Porous Metals.
NDT& E International 39, no. 7, pp. 5628. doi: http://dx.doi.org/10.1016/j.
ndteint.2006.03.008
Ma, X., A.J. Peyton, S.R. Higson, A. Lyons, and S.J. Dickinson. 2006. Hard-
ware and Software Design for an Electromagnetic Induction Tomography
(EMT) System Applied to High Contrast Metal Process Applications. Mea-
surement Science and Technology 17, no. 1, pp. 11118. doi: http://dx.doi.
org/10.1088/0957-0233/17/1/018
Ma, X., A.J. Peyton, S.R. Higson, and P. Drake. 2008. Development of Mul-
tiple Frequency Electromagnetic Induction Systems for Steel Flow Visual-
ization. Measurement Science and Technology 19, no. 9. doi: http://dx.doi.
org/10.1088/0957-0233/19/9/094008
Ma, X. 2012. Investigations of the State-of-the-Art Methods for Electromag-
netic NDT and Electrical Condition Monitoring. Insight-Non-Destructive
Testing and Monitoring 54, no. 9, pp. 4828. doi: http://dx.doi.org/10.1784/
insi.2012.54.9.482
Ma, H., J.C. Chan, T.K. Saha, and C. Ekanayake. 2013. Pattern Recognition
Techniques and Their Applications for Automatic Classification of Arti-
ficial Partial Discharge Sources. IEEE Transactions on Dielectrics and
Electrical Insulation 20, no. 2, pp. 46878. doi: http://dx.doi.org/10.1109/
tdei.2013.6508749
Markalous, S., S. Tenbohlen, and K. Feser. 2008. Detection and Location of Par-
tial Discharges in Power Transformers using Acoustic and Electromagnetic
Signals. IEEE Transactions on Dielectrics and Electrical Insulation 15,
no. 6, pp. 157683. doi: http://dx.doi.org/10.1109/tdei.2008.4712660
Monitoring Degradation in the Field 149
Marsh, L.A., C. Ktistis, A. Jarvi, D.W. Armitage, and A.J. Peyton. March 2014.
Three-dimensional Object Location and Inversion of the Magnetic Polar-
izability Tensor at a Single Frequency Using a Walk-through Metal Detec-
tor. Measurement Science and Technology 24, no. 4. doi: http://dx.doi.
org/10.1088/0957-0233/24/4/045102
Mayoux, C. October 2000. Degradation of Insulating Materials Under Electrical
Stress. IEEE Transactions on Dielectrics and Electrical Insulation 7, no. 5,
pp. 590601. doi: http://dx.doi.org/10.1109/tdei.2000.879355
Mazzetti, C., F.M.F. Mascioli, F. Baldini, M. Panella, R. Risica, and R. Bartni-
kas. 2006. Partial Discharge Pattern Recognition by Neuro-fuzzy Networks
in Heat-shrinkable Joints and Terminations of XLPE Insulated Distribution
Cables. IEEE Transactions on Power Delivery 21, no. 3, pp.103544. doi:
http://dx.doi.org/10.1109/tpwrd.2006.875861
Mercier, D., J. Lesage, X. Decoopman, and D. Chicot. December 2006. Eddy
Currents and Hardness Testing for Evaluation of Steel Decarburizing.
NDT& E International 39, no. 8, pp. 65260. doi: http://dx.doi.org/10.1016/j.
ndteint.2006.04.005
Mook, G., R. Lange, and O. Koeser. May 2001. Non-destructive Characterisation
of Carbon-fibre-reinforced Plastics by Means of Eddy-currents. Composites
Science and Technology 61, no. 6, pp. 86573. doi: http://dx.doi.org/10.1016/
s0266-3538(00)00164-0
Pearson, J.S., O. Farish, B.F. Hampton, M.D. Judd, D. Templeton, B.M. Pryor, and
I.M. Welch. October 1995. Partial Discharge Diagnostics for Gas Insulated
Substations. IEEE Transactions on Dielectrics and Electrical Insulation 2,
no. 5, pp. 893905. doi: http://dx.doi.org/10.1109/94.469984
Riedel C.H., M. Keppelen, S. Nani, R.D. Merges, and O. Dssel. 2004. Planar
System for Magnetic Induction Conductivity Measurement using a Sensor
Matrix. Physiological Measurement 25, no. 1, pp. 40311. doi: http://dx.doi.
org/10.1088/0967-3334/25/1/043
Rogers, R.R. 1978. Concepts used in the Development of the IEEE and IEC
Codes for the Interpretation of Incipient Faults in Power Transformers by Dis-
solved Gas in Oil Analysis. IEEE Winter Meeting EI-13, no. 5. doi: http://
dx.doi.org/10.1109/TEI.1978.298141
Rubio-Serrano, J., M.V. Rojas-Moreno, J. Posada, J.M. Martinez-Tarifa,
G. Robles, and J.A. Garcia-Souto. 2012. Electro-acoustic Detection,
Identification and Location of Partial Discharge Sources in Oil-paper
Insulation Systems. IEEE Transactions on Dielectrics and Electrical
Insulation 19, no. 5, pp.156978. doi: http://dx.doi.org/10.1109/
tdei.2012.6311502
Rudd, S., S.D.J. Mcarthur, and M.D. Judd. 2010. A Generic Knowledge-based
Approach to the Analysis of Partial Discharge Data. IEEE Transactions on
Dielectrics and Electrical Insulation 17, no. 1, pp.14956. doi: http://dx.doi.
org/10.1109/tdei.2010.5412013
Sher, Z.K., D. Zhu, X. Jin, and K. Tan. 1995. A New Adaptive Technique
for On-line Partial Discharge Monitoring. IEEE Transactions on Dielec-
150THE PHYSICS OF DEGRADATION
Physics of Degradation
in Ferroelectric Devices
Paul M. Weaver
6.1INTRODUCTION
DOI: 10.5643/9781606504673/ch6
154THE PHYSICS OF DEGRADATION
P
Pmax
Ps
Pr
E
Ec
ele
Piezo ctric
Pyroelectric
rroelectri
Fe
c
6.2HUMIDITY
the material subject to the electric field. These sites appear to be related
to pores, and it is likely that pores in the ceramic provide a route for
ingress of moisture. This is supported by the observation that high-den-
sity, low-porosity ceramics show significantly longer lifetime in humid
conditions. However, pores are not the only mechanism for water vapor to
affect the ceramic. Conduction also occurs at edges (and surfaces exposed
by cracking). Edge conduction becomes a more prominent contributor in
low-porosity ceramics. In this case, surface contamination and poor sur-
face condition can significantly increase the rate of growth of the leakage
current. These observations suggest that porosity is a major factor in the
development of leakage currents in ceramics exposed to high humidity
and high electric fields, although electrode material and processing condi-
tions also have an effect.
Polymer barrier coatings provide protection against gross contamina-
tion from water drops and condensation, but provide little impediment to
the passage of water vapor from the air. They are therefore generally inef-
fective in preventing long-term protection against the effects of humidity.
Combination coatings such as metal or diamond-like carbon (DLC) on a
polymer base may extend lifetime in humid conditions, but processing
to achieve full-area coverage and sealing at electrical connections pres-
ent significant challenges. Control over the manufacturing processes from
ceramic through to device integration is crucial for achieving cost-effec-
tive performance in harsh environments.
confers high electrical resistivity (B. Jaffe, Cook, and H. Jaffe 1971). Data
from B. Jaffe, Cook, and H. Jaffe (1971) would indicate a volume resistiv-
ity of 109 cm at temperatures around room temperature. For the sample
size and field strength used in this study we would expect a resistance
of over 300 G, and a leakage current of less than 1 nA. The currents
measured after exposure to humid conditions under d.c. bias are in the
A to mA range, so it cannot be explained by these electronic conduction
mechanisms. In addition, there is no obvious reason why bulk conduction
mechanisms should be directly affected by humidity, or the duration of the
stress condition (humidity + d.c. field), although secondary effects such as
humidity-induced chemical changes could play a part.
There is evidence that the conduction mechanism in humid envi-
ronments is ionic in character. This includes the observation of anodic
breakdown features and sensitivity of the leakage current to the relative
humidity and electrode material (Lipscomb 2009). The large leakage cur-
rents observed under humid conditions are not observed under dry atmo-
sphere (<250 nA at 55C), confirming that the conductivity is caused by
the humidity. In fact, a reduction of leakage current is usually observed on
heating to this temperature from room temperature due to the expulsion
of moisture. The effects of the bulk conduction mechanisms referred to
above are only easily measurable above around 200C.
Humidity ingress has long been recognized as a potential failure
mode in ceramic capacitors, which are often based on ferroelectric mate-
rials such as barium titanate (BaTiO3) and operate under similar levels
of electric field to ferroelectrics in other applications such as actuators.
There have been a number of studies of the effects of DC fields on elec-
trical breakdown in multilayer capacitors (MLCs). Waser, Baiatu, and
Hardtl (1990) have termed resistance degradation as a limiting factor in
the lifetime of MLCs. In MLCs, the effects of humidity have also been
investigated, particularly with silver electrodes (Ling and Jackson 1989).
Subsequent studies on the reliability of multilayer ceramic actuators have
also examined the behavior of silver electrodes under d.c. operation in
humid conditions (Nagata and Thongrueng 1998; Pritchard, Bowen, and
Lowrie 2001; Thongrueng, Tsuchiya, and Nagata 1998). More recently,
the design of multilayer actuators has been optimized to reduce the effects
of humidity by burying the electrode within the actuator, and through the
use of protective ceramic coatings (Bindig and Helke 2000; Pertsch et al.
2006).
More recently, the effects of water vapor have been studied in the con-
text of memory and electronic applications Baniecki et al. (2002) where it
has been shown to have a strong influence on the leakage characteristics of
PZT thin films. The application of large d.c. electric fields to piezoelectric
158THE PHYSICS OF DEGRADATION
Electrode +
4. Breakdown
zone
Ceramic
2. Cation
migration
3. Filament
growth
Electrode
0.2
0.18
0.16
0.14
Ngcmcig"ewttgpv"*oC+
0.12
0.1
0.08
0.06
0.04
0.02
0
0 5 10 15 20 25 30 35 40 45
Vguv"vkog"*jqwtu+
Hole Electrode
Molten
area
20 m
20 m
damage to the ceramic structure. The resistivity of the thin electrode mate-
rial (approximately 2 m) will serve to limit the overall current, although
a localized high-current density on breakdown is likely. It is thought that
this current-limited breakdown transient is responsible for the current fluc-
tuations (Zheng, Swingler, and Weaver 2010). The fact that these features
only occur on the anode is strong evidence for the ionic migration/metal
filament mechanism for the growth of the conductive channels.
At the center of the fused area is a hole. It is possible that this was
originally a surface pore that provided a location for entry of the water
vapor required to promote electromigration. This observation supports the
idea that pores play an important role in the susceptibility to humidity and
that the porosity of the ceramic is an important factor.
a test showed that the leakage current was rapidly restored to its value
before removal, confirming that the edges were not a major contributor to
the conductivity (Weaver et al. 2012). In another experiment, a ceramic
sample was tested for approximately 20 hours, then removed from test and
broken into two approximately equal halves. Both halves were retested
separately. After a short time for recovery of the leakage current, the two
halves showed approximately equal leakage currents that were approx-
imately half the current before breaking in two. The breaking of the
ceramic pieces creates two new edges with a combined length of approxi-
mately 33 percent of the original perimeter. This increase in edge area was
not reflected in a corresponding increase in leakage current. There is evi-
dence, however, that edge effects become significant at longer timescales
in ceramics where the leakage by other mechanisms has been suppressed.
(a) (b)
50 m
50 m
(a) (b)
50 m 50 m
Figure 6.8. SEM cross sections showing (a) porosity in the standard ceramic
and (b) low-porosity version of the ceramic (Weaver et al. 2012).
SEM cross sections of two ceramics of the same composition with differ-
ent porosity.
Test results demonstrated very significant reductions in leakage current
in the high-porosity version of the ceramic. The longer lifetimes achieved
with low-porosity ceramics means that factors, such as edge effects, that did
not play a significant role in conduction through previously now become a
limiting factor. Optical microscope study of a tested low-porosity ceramic
(after electrode removal) showed that eventually discolorations of the
ceramic due to breakdown events were observed, but that these were mostly
located at the edges rather than through the bulk. It is thought that this edge
conduction can ultimately lead to short circuit failure but at much longer
timescales. Contamination was also found to play a role in the conduction
processes in low-porosity ceramics that was masked in the more porous
ceramics. Aqueous cleaning to remove ionic contaminants and careful han-
dling was found to improve lifetime and consistency in performance.
Electrode materials and processing conditions affect the level of leakage cur-
rent observed. As conduction occurs through the bulk of the material away
from the edges, it must be able to penetrate the metallic electrode material
through coverage defects, cracks, or incompletely covered surface pores.
This would imply that thicker electrodes may improve lifetime in humidity.
However, this approach increases the stiffness of the device possibly leading
to unacceptable degradation of device performance, particularly for thinner
ceramics. Gold electrodes have been demonstrated to significantly improve
performance (Lipscomb et al. 2009). However, Au electrodes can be expen-
sive particularly if thickness is required to overcome porosity.
164THE PHYSICS OF DEGRADATION
Actuator
Open
movement
Actuator Closed
Valve opening
be sufficient clearance between the valve seal and the orifice to achieve
the required flow rate. In the closed position, the valve seal needs to at
least cover the orifice and, usually, some overtravel is needed to create
sufficient force to create an effective seal. These two positions define the
range of movement required from the actuator between the ON condition
and the OFF condition. Note that these two positions are relative to the
valve orifice. If there is any drift in the position of the actuator, then it
could encroach on these limits and compromise the performance of the
valve. This is why thermal expansion is an important factor in the design
and reliability of many piezoelectric devices. In fact, it is the difference
in thermal expansion between the piezoelectric material and the support-
ing structure that leads to thermal drift. For example, a typical actuation
strain may be around 1,000 ppm. A thermal expansion difference of 10
ppm would create a drift equal to the entire actuation strain over a 100C
temperature span. Such a valve would only work across a limited tem-
perature range.
The piezoelectric materials most widely used for actuator applica-
tions such as the valve are ceramics of ferroelectric materials, usually
lead zirconate titanate (PZT), because of the very coupling between their
electrical polarization and mechanical strain. As described above, these
materials are also pyroelectric, with strong coupling between electrical
166THE PHYSICS OF DEGRADATION
(el
ctr cal
ect
Electrical
ele ani
ic)
Ele loric/
roc
e z o ch
ctr
(pi rome
a
oth yroe
ct
erm lec
p
Ele
al tric)
Mechanical Thermal
Thermoelastic
(expansion/piezocaloric)
S = QP 2 (6.1)
P = c E (6.2)
Sr
P
Pr
(a)
100
50
0
-S (ppm)
50
100
150
S0
200
0.2 0.15 0.1 0.05 0 0.05 0.1 0.15 0.2
P (Cm2)
(b)
1600
1400
1200
1000
800
-S (ppm)
600
400
200
0 S0
200
0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4
P (Cm2)
1400
1200 19C
1000 49C
800
/U"*rro+
77C
600 104C
400 137C
200 167C
186C
0
200
0 20 40 60 80 100 120 140 160 180 200
Vgorgtcvwtg"*E+ Rqnctk|cvkqp"*207"Eo4"rgt"fkxkukqp+
1600
a
1400
1200
1000
800
/U"*rro+
600
400 b
200
0 c
200
d
400
0 50 100 150 200
Vgorgtcvwtg"*E+
S (T , P) = S0 (T ) + QP 2 (1.3)
170THE PHYSICS OF DEGRADATION
c+
0.012
0.010
0.008
/S34"*E4o6+
0.006
0.004
0.002
0.000
0 20 40 60 80 100 120 140 160 180 200
V"*E+
d+
0
20 1.33 ppm C1
40 1.77 ppm C1
60
/U2"*rro+
80
100
120
140
160
180
0 20 40 60 80 100 120 140 160 180 200
V"*E+
Figure 6.15. Electrostriction coefficient, (a) Q12, and (b) strain zero S0 as
functions of temperature (bipolar cycles). In (b) the dashed lines represent
straight line fits to the approximately linear portions of the curves.
Piezoelectric ceramic
Inactive substrate
Figure 6.16.Piezoelectric
unimorph actuator operating
principle.
172THE PHYSICS OF DEGRADATION
0.6
a
0.4
0.2
Rqukvkqp"*oo+ b
0
0.2
c
0.4
0.6
0.8 d
1
40 20 0 20 40 60 80 100
Vgorgtcvwtg"*E+
low-temperature stroke would be the limiting factor in the design of, for
instance, a valve actuator. Application of an electric field opposing the
polarization direction can remove some of this build-up of remanence at
low temperature. Figure 6.18 shows how such a reverse field affects the
strain in the ceramic and the actuator position.
A moderate reverse field can significantly increase the movement,
particularly where the remanence is large. However, too high an electric
field exceeds the coercive voltage, resulting in semi-permanent repolar-
ization of the actuator causing it to drive in a direction opposite to that
intended. Because the coercive field is strongly temperature dependent
(Figure 6.19), applying a fixed reverse field would only work over a fairly
restricted temperature range.
Whats needed is some kind of adjustable reverse voltage that takes
account of the temperature variation of the coercive field to avoid repoling
the ceramic. Such a system based on charge measurements is described in
Weaver (2011). The application of a temperature-varying reverse field is
shown in curve (c) in Figure 6.17 showing a significant reduction in the
remanence without repoling at high temperature. Even after the reverse
field has been removed, there is still some benefit as shown in curve (d) in
Figure 6.17. This technique results in a significant improvement in both
stroke and operating window for the actuator.
Physics of Degradation in Ferroelectric Devices 173
d
Discharge
Reverse
charge
Ec +Ec E
0.6
0.5
Eqgtekxg"hkgnf"*mX"oo3+
0.4
0.3
0.2
0.1
0
0 20 40 60 80 100 120 140 160 180 200
Vgorgtcvwtg"*E+
Base
Mirrored glass
cantilever
6.5CONCLUSION
ACKNOWLEDGMENTS
REFERENCES
Baniecki, J.D., J.S. Cross, M. Tsukada, and J. Watanabe. 2002. H2O Vapor-In-
duced Leakage Degradation of Pb(Zr,Ti)O3 Thin-Film Capacitors with Pt and
IrO2 Electrodes. Applied Physics Letters 81, no. 20, pp. 383739. doi: http://
dx.doi.org/10.1063/1.1519359
Bindig, R. and G. Helke. 2000. Application of piezoceramic multilayer actuators,
experiences and solutions. Proceedings on Actuator 2000, pp. 537. Bremen,
Germany.
Cain, M. 2014. Dielectric Breakdown in Dielectrics and Ferroelectric Ceram-
ics. In Characterisation of Ferroelectric Bulk Materials and Thin Films,
pp. 24366. Springer.
Chen, W.P., H.L.W. Chan, F.C.H. Yiu, K.M.W. Ng, and P.C.K. Liu. 2002.
Water-Induced Degradation in Lead Zirconate Titanate Piezoelectric Ceram-
ics. Applied Physics Letters 80, no. 19, pp. 358789. doi: http://dx.doi.
org/10.1063/1.1479205
Chen, W.P., X.P. Jiang, Y. Wang, Z. Peng, and H.L.W. Chan. 2003. Water-Induced
Degradation of Barium Titanate Ceramics Studied by Electrochemical
Hydrogen Charging. Journal of the American Ceramic Society 86, no. 4,
pp. 73537. doi: http://dx.doi.org/10.1111/j.1151-2916.2003.tb03366.x
Dawber, M., K. Rabe, and J. Scott. October 2005. Physics of Thin-Film Ferro-
electric Oxides. Reviews of Modern Physics 77, no. 4, pp. 1083130. doi:
http://dx.doi.org/10.1103/revmodphys.77.1083
Huang, Z. and G. Leighton. 2014. Interferometry for Piezoelectric Materials and
Thin Films. In Characterisation of Ferroelectric Bulk Materials and Thin
Films, ed. M.G. Cain, 1536, Springer.
Jaffe, B., W. Cook, and H. Jaffe. 1971. Piezoelectric Ceramics. London: Academic
Press.
Jiang, B., Y. Bai, J.-L. Cao, Y. Su, S.-Q. Shi, W. Chu, and L. Qiao. 2008. Delayed
Crack Propagation in Barium Titanate Single Crystals in Humid Air. Journal of
Applied Physics 103, no. 11, p. 116102. doi: http://dx.doi.org/10.1063/1.2921136
Jones, B.J., A. Mahendran, A.W. Anson, A.J. Reynolds, R. Bulpett, and J. Franks.
2010. Diamond-Like Carbon Coating of Alternative Metal Alloys for Medi-
cal and Surgical Applications. Diamond and Related Materials 19, no. 79,
pp. 6859. doi: http://dx.doi.org/10.1016/j.diamond.2010.02.012
178THE PHYSICS OF DEGRADATION
Kostic, A., A. Rensch, and D. Sturm. 1995. Nickel Dendrites: A New Failure
Mechanism in Ceramic Hermetic Packages. In Reliability and Maintainabil-
ity Symposium, 1995. Proceedings, Annual, 3943. Washington DC: IEEE.
Ling, H.C. and A.M. Jackson. March 1989. Correlation of Silver Migration with
Temperature-Humidity-Bias (THB) Failures in Multilayer Ceramic Capaci-
tors. IEEE Transactions on Components, Hybrids, and Manufacturing Tech-
nology 12, no. 1, pp. 1307. doi: http://dx.doi.org/10.1109/33.19027
Lipscomb, I.P., P.M. Weaver, J. Swingler, and J.W. McBride. 2008. Micro-
Computer TomographyAn Aid in the Investigation of Structural Changes in
Lead Zirconate Titanate Ceramics after Temperature-Humidity Bias Testing.
Journal of Electroceramics 23, no. 1, pp. 725. doi: http://dx.doi.org/10.1007/
s10832-008-9537-8
Lipscomb, I.P., P.M. Weaver, J. Swingler, and J.W. McBride. 2009. The Effect of
Relative Humidity, Temperature and Electrical Field on Leakage Currents in
Piezo-Ceramic Actuators under DC Bias. Sensors and Actuators A: Physical
151, pp. 17986. doi: http://dx.doi.org/10.1016/j.sna.2009.01.017
Lupascu, D.C. 2004. Fatigue in Ferroelectric Ceramics and Related Issues,
Vol.61. Heidelberg: Springer Verlag.
Massey, L.K. 2003. Permeability Properties of Plastics and ElastomersA Guide
to Packaging and Barrier Materials. 2nd ed. William Andrew Publishing.
McCartney, L., L. Wright, M. Cain, J. Crain, G. Martyna, and D. Newns. 2014.
Methods for Determining Piezoelectric Properties of Thin Epitaxial Films:
Theoretical Foundations. Journal of Applied Physics 116, no. 1, p. 014104.
doi: http://dx.doi.org/10.1063/1.4885058
Mock, R. and K. Lubitz. 2008. Piezoelectric Injection Systems. Piezoelectric-
ity, Springer Series in Materials Science 114, pp. 299310. doi: http://dx.doi.
org/10.1007/978-3-540-68683-5_13
Nagaraj, B., S. Aggarwal, T. Song, T. Sawhney, and R. Ramesh. 1999. Leak-
age Current Mechanisms in Lead-Based Thin-Film Ferroelectric Capacitors.
Physical Review B 59, no. 24, pp. 160227. doi: http://dx.doi.org/10.1103/
PhysRevB.59.16022
Nagata, K. and J. Thongrueng. 1998. Effect of Temperature, Humidity and Load
on Degradation of Multilayer Ceramic Actuator. Journal of the Korean Phys-
ical Society 32, pp. 127881.
Pertsch, P., S. Richter, D. Kopsch, N. Kramer, J. Pogodzik, and E. Henning. 2006.
Reliability of piezoelectric multilayer actuators. In Actuator 2006, pp. 527
30. Bremen, Germany.
Pritchard, J., C. Bowen, and F. Lowrie. 2001. Multilayer Actuators: Review.
British Ceramic Transactions 100, no. 6, pp. 26573. doi: http://dx.doi.
org/10.1179/096797801681549
Rusek, K., J. Kruczek, K. Szot, D. Rytz, M. Gorny, and K. Roleder. 2008.
Non-Linear Properties of BaTiO3 above T C. Ferroelectrics 375, no. 1,
pp. 1659.
Schmitz-Kempen, T., S. Tiedke, P. Mardilovich, S. Sivaramakrishnan,
T.Lisec, F. Stoppel, S. Trolier-McKinstry, and P. Muralt. 2013. Compa-
Physics of Degradation in Ferroelectric Devices 179
Dr. Jonathan Swingler received a Joint BSc (Hons) in Physics and Chem-
istry from Keele University in 1990 followed by a PhD at Loughborough
University for his work on the degradation of electrical contacts. He sub-
sequently moved to the University of Southampton where he pursued
his research into the physics of degradation and reliability of electrical
and electronic materials and devices. Currently Jonathan is concerned to
develop reliability engineering science at Heriot-Watt University to enable
the prediction of electrical component and system reliability under harsh
operating conditions. He is an Associate Professor of Energy at Heri-
ot-Watt University and Fellow of the Institute of Physics.
Dr. Jeff Jones is an Associate Professor at WMG with research inter-
ests in technical asset management and dependability. The focus of his
research is the use of models, simulations, and artificial intelligence, and
these have been applied to large multifaceted systems down to MEMS
devices. He is the chair of IEC TC/56 WG2 on Dependability Techniques,
and is principal UK expert on a number of drafting committee within IEC.
He has recently been awarded the 1906 award by IEC for his work in
standardization.
Dr. Michael D. Bryant, Accenture Endowed Professor of Mechanical
Engineering, specializes in tribology, the design of mechatronic systems
and system diagnostics of machinery. Dr. Bryant is a faculty member of
the Dynamic Systems & Control and Manufacturing & Design programs
of the Mechanical Engineering Department. Dr. Bryant earned his Ph.D.
in engineering sciences and applied mathematics from Northwestern Uni-
versity in 1981. He has served on the Cockrell School of Engineering fac-
ulty since 1988 and previously taught at North Carolina State University.
He has published more than 100 technical articles and reports. In 2005,
he was appointed Editor in Chief of the ASME Journal of Tribology and
Series Editor for Dynamic Systems & Control and Mechatronics, Springer
Mechanical Engineering Series. He is a Fellow of ASME and member
or IEEE.
182 About the Contributing Authors
H N
H-shaped electromagnetic sensor, NDT. See Nondestructive testing
122123 (NDT)
High-performance liquid Nernst thermodynamic equilibrium
chromatography (HPLC), condition, 89
132 Non-equilibrium thermodynamic
damage assessment, 67
I cyclic cumulative damage
Insulation degradation acceleration factors, 8083
dielectric discharge model corrosion current, 8889
corona discharges, 128 corrosion rate, 90
internal discharges, 128 cyclic reversible work, 76
surface discharges, 128 cyclic work plane, 77
three-capacitor model, electrochemistry, 8488
129130 Miners rule (see Miners
voltages and current pulses, rule)
128129 Nernst thermodynamic
multiple physical stresses equilibrium
localized electrical stress, condition, 89
128 non-reversible cyclic work,
mechanical stress, 127128 7677
thermal ageing, 127 Stokes integral theorem, 78
partial discharge (see thermodynamic damage
Partial discharge (PD) ratio method,
detection techniques) 7880
Interferometry, 174 vs. equilibrium
thermodynamics
L damage
Logarithmic-in-time ageing TAT conjugate work, 76
model, 100101 free energy approach, 76
quasi-static measurement,
M 75
Miners rule mechanical systems
cyclic damage, 8183 creep damage, 9193
modification, 8384 thermal cycle fatigue,
secondary batteries 9596
activation free energy, 85 vibration cyclic fatigue
chargedischarge cycles, damage, 9698
84 wear damage, 9394
Index 187
Industrial Engineering
Industrial, Systems, and Innovation Engineering William R. Peterson,CollectionEditor
Manufacturing and Processes Wayne Hung, Collection Editor
General Engineering and K-12 Engineering Education John K. Estell and Kenneth
J. Reid, Collection Editors
Electrical Engineering
E lectrical Power Hemchandra M. Shertukde, Collection Editor
Communications and Signal Processing Orlando Baiocchi, CollectionEditor
Computer Engineering Foundations, Currents and Trajectories Augustus (Gus) Kinzel
Uht, Collection Editor
Electronic Circuits and Semiconductor Devices Ashok Goel, Collection Editor
Civil Engineering
Environmental Engineering Francis Hopcroft, Collection Editor
Geotechnical Engineering Hiroshan Hettiarachchi, Collection Editor
Transportation Engineering Bryan Katz, Collection Editor
Sustainable Structural Systems Mohammad Noori, Collection Editor
Momentum Press actively seeking collection editors as well as authors. For more
information about becoming an MP author or collection editor, please visit
http://www.momentumpress.net/contact
The Momentum Press digital library is very affordable, with no obligation to buy in future years.
For more information, please visit www.momentumpress.net/library or to set up a trial in the US,
please contact mpsales@globalepress.com.
EBOOKS The Physics of Degradation in MATERIALS PROPERTIES AND
SWINGLER
FOR THE Engineered Materials and Devices BEHAVIOR COLLECTION
ENGINEERING
Fundamentals and Principles
LIBRARY
Edited by Jonathan Swingler with major
Create your own
contribution by Alec Feinberg
Customized Content
Bundlethe more Degradation is apparent in all things and is fundamental to both
THE CONTENT
Manufacturing
key players engaged in the physics of degradation to present
the background science, current thinking and developments in
understanding, and gives a detailed account of emerging issues
Degradation in
Engineered Materials
across a selection of engineering applications. The work has been
Engineering
put together to equip the upper level undergraduate student,
Mechanical
postgraduate student, as well as the professional engineer and
& Chemical
and Devices
scientist, in the importance of physics of degradation. The aim
Engineering of The Physics of Degradation in Engineered Materials and
Materials Science Devices is to bridge the gap between published textbooks
& Engineering on the fundamental science of degradation phenomena
Civil & and published research on the engineering science of actual
Environmental
Engineering
fabricated materials and devices.
A history of the observation and understanding of physics of Fundamentals and Principles
Advanced Energy degradation is presented and the fundamentals and principles
Technologies of thermodynamics and entropy are extensively discussed. This
is the focus of this book, with an extended chapter by Alec
Feinberg on equilibrium thermodynamic damage and non-
THE TERMS equilibrium thermodynamic damage. It concludes with two
Perpetual access for particular technologies to give examples of areas of application.
a one time fee
No subscriptions or Dr. Jonathan Swingler received a Joint BSc (Hons) in physics
access fees and chemistry from Keele University in 1990, followed by a PhD
at Loughborough University for his work on the degradation of
Unlimited
electrical contacts. He subsequently moved to the University of
concurrent usage
Southampton where he pursued his research into the physics of
Downloadable PDFs
degradation and reliability of electrical and electronic materials
Free MARC records and devices. Currently Jonathan is working to develop reliability-
engineering science at Heriot-Watt University, to enable the
Edited by Jonathan Swingler
For further information,
a free trial, or to order,
contact:
prediction of electrical component and system reliability under
harsh operating conditions. He is an associate professor of with major contribution
energy at Heriot-Watt University and fellow of the Institute of
sales@momentumpress.net Physics. by Alec Feinberg
ISBN: 978-1-60650-467-3