Design for Reliability of Power Electronic Systems

Huai Wang, IEEE Member, Ke Ma, IEEE Student Member, Frede Blaabjerg, IEEE Fellow
Center Of Reliable Power Electronics (CORPE)
Department of Energy Technology, Aalborg University
Pontoppidanstraede 101, DK-9220, Aalborg East, Denmark
[email protected]; [email protected]; [email protected]

Abstract - Advances in power electronics enable efficient and Small, Group I Medium, Group II Large, Group III
'.0
flexible processing of electric power in the application of
renewable energy sources, electric vehicles, adjustable-speed J5

drives, etc. More and more efforts are devoted to better power .� 3.0
electronic systems in terms of reliability to ensure high
2.5
..
'7

availability, long lifetime, sufficient robustness, low maintenance

:e
C
2.0
cost and low cost of energy. However, the reliability predictions
!
.a
15

�
are still dominantly according to outdated models and terms,
"

Failure (MTTF) , and Mean-Time-Between-Failures (MTBF). A

such as MJL-HDBK-217H handbook models, Mean-Time-To­ 1.0

• •
(PoF)
0.5
collection of methodologies based on Physics-of-Failure
• •
;:!�
� "'.-'� U>U>�8��
0.0

�� �� ::�
� §� � ��
approach and mission profile analysis are presented in this

sf§
>'- 0

!8 !8 �
i� z� �� �� � ��
0 >'-
z
paper to perform reliability-oriented design of power electronic 0� ,.�
��
.!!�
�
�

;; �� ;; � 8 ��
-N
��
o
�"'

:N 3�
C o

:il
0 0
systems. The corresponding design procedures and reliability >
0", 0U>

" > �
� .

� 'E� ��
GO U
Z �
C o
prediction models are provided. Further on, a case study on a o �
"I
z
0_

«
C
2.3 MW wind power converter is discussed with emphasis on the
reliability critical components JGBTs. Different aspects of
Fig. 1. Dependence of failure rate on power rating of wind turbines [4].
improving the reliability of the power converter are mapped.
Finally, the challenges and opportunities
reliable power electronic systems are addressed.

Index Terms
to achieve more

:· · · ·� · · · �cP:::
: i:· · · · · · · i:· @· ·OH�
· · · · i:· ·- -� - - - -!O�
:- - - - - - -·i
: : : : : : : : :
Design for reliability, power electronics, wind I , , I , I I • •
-

: : : : : :
I , I , I , , I •

: : :
--- - -- - - - -
, , , , , , , I •
I , , , , , , , •
power converter, JGBT modules, physics-of-failure
-- - ------ -- - --- - -- ---------- - -------------. .----------.-.

J!l Turbines Gearbox Generator Electrical Control

'"
� 1/2
I. INTRODUCTION � Hydraulic Blades
� 1/4
Power electronics have enabled efficient conversion and �

§�
Ql
more flexible control of electric energy in the last four 2
decades. However, the reliability performance of power �D-
� 4
o
electronic systems imposes huge challenges in various
emerging applications, especially for the grid integration of
renewable energy with long operation hours under harsh Fig. 2. Distribution of failure rate and downtime among various parts of the
wind turbines in [5].
environment. [t has significant impact on the life cycle cost of
the systems, levelized cost of energy and customer
the larger WTs are prone to have higher failure rate as shown
satisfaction, therefore, the penetration of renewable energy in
in Fig. 1 [4], which is according to the failure statistics on all
our modem electrical grid is a challenge in the long run.
of the WTs operating during 1993 -2006 in Schleswig­
[n wind power generation system, power electronic
Holstein, Germany. Regarding the failure rate distribution
converters are dominantly applied for regulating the
within a certain type of WT, an analysis is given in [5] and
fluctuating input power and maximizing the electrical energy
represented in Fig. 2. It can be noted that power converters
harvested from the wind [1]- [2]. The penetration of wind
dominant the failures, which is in line with the observations
power is expected to be 20% of the total electricity production
from other field surveys in [6]-[7].
by 2020 in Europe [3]. Meanwhile, the power capacity of a
In photovoltaic (PV) system, PV inverters are used for
single wind turbine (WT) is increasing from tens of kW to 10
efficiently converting the dc voltage for ac applications or
MW and the location of the wind farm is moving from
integration of the output energy into electrical grid [8]. With
onshore to offshore. Therefore, the reliability performance of
the progressive development of PV around the world, PV
the whole system is of primary concern due to increased time
inverters are becoming the most critical subsystems in terms
and cost for repairs after failures, having significant impact on
of failure rate, lifetime and maintenance cost. Leading
the availability of the wind power generation. Unfortunately,
manufacturers nowadays provide PV modules with over 20

978-1-4673-2421-2/12/$31.00 ©2012 IEEE 33

Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on January 27,2025 at 12:53:26 UTC from IEEE Xplore. Restrictions apply.
years of warranty. However, the number is around 5 years for c) Over reliance on handbook-based models and
PV inverters on average in 2012 [9]. Therefore, even though statistics. Military handbook MIL-HDBK-217F [39] is widely
inverters account only for 10%-20% of the initial system cost, used to predict the failure rate of power electronic components
they could need to be replaced 3 -5 times over the life of a PV [22], [26]- [27] and [29]. However, temperature cycling,
system, introducing additional investment [10]. According to failure rate change with material, combined environments,
the 5 years of field experience in a large utility-scale PV supplier variations (e.g. technology and quality) are not
generation plant studied in [ll], the PV inverters are considered. Moreover, as failure details are not collected and
responsible for 3 7% of the unscheduled maintenance and 59% addressed, the handbook method could not give designers any
of the associated cost. insight into the root cause of a failure and the inspiration for
Industries have advanced the development of reliability reliability enhancement. Physically, a failure rate of a
engineering from traditional testing for reliability to Design component is the sum of the failure rates of all failure modes,
For Reliability (DFR) [12]. DFR is the process conducted which have different reliability models corresponding to
during the design phase of a component or system that ensures specific failure mechanisms. Statistics is a necessary basis to
them to be able to perform required level of reliability. It aims deal with the effects of uncertainty and variability on
to understand and fix the reliability problems up-front in the reliability. However, as the variation is often a function of
design process. Accordingly, many efforts have been devoted time and operating conditions, statistics itself is not sufficient
to considerations into the reliability aspect performance of to interpret the reliability data without judgment of the
power electronic components [13]- [20], converters [21]-[29] assumptions and non-statistical factors (e.g. modification of
and systems [30]- [35]. However, the reliability research in design, new generation of components, etc.).
the power electronics area has the following limitations: Therefore, the scope of this paper is first to give an
a) Lack of systematic DFR approach specific for design introduction of failures in power electronic systems. Then a
of power electronic systems. The DFR approach studied in systematic DFR procedure based on physics-of-failure (PoF)
reliability engineering is too broad in focus [12]. Power [40] approach and mission profile analysis is proposed in
electronic systems have their own challenges and new section III. A case study on a 2.3 MW wind power converter
opportunities in enhancing the reliability, which is worthwhile is presented in section IV with emphasis on the reliability
to be investigated. Moreover, design tools, except for the criticallGBT modules and the potential methods for reliability
reliability prediction, are rarely applied in state-of-the-art enhancement. The challenges and opportunities are addressed
research on reliability of power electronic systems. in the conclusions.
b) Over reliance on calculated value of Mean-Time-To­
Failure (MTTF) or Mean-Time-Between-Failures (MTBF) II. FAILURES IN POWER ELECTRONIC SYSTEMS
and bathtub curve [36]. Bathtub curve divides the operation of Reliability is defined as the ability of an item to perform
a device or system into three distinct time periods. Although it required function under stated conditions for a certain period
is approximately consistent with some practical cases, the of time, which is often measured by probability of failure,
assumptions of "random" failure and constant failure rate frequency of failure, or in terms of availability. The essence
during the useful life period are misleading [36] and the true of reliability engineering is to prevent the creation of failures.
root causes of different failure modes are not identified. The Deficiencies in the design phase have effect on all produced
fundamental assumptions of MTTF or MTBF are constant items and the cost to correct them is progressively increased as
failure rate and no wear out. Therefore, the calculated values development proceeds. Therefore, this section introduces the
may have high degree of inaccuracy if wear out occurs within following aspects of failure in power electronic systems.
the time. For example, based on the assumptions, modern PV
modules could have MTBF between 500 and 6000 years [3 7], A. Failure Criteria
however, lifetime and warranty time (e.g. 20 years) are much In power electronic systems, the degradation of one
less than this due to wear out failures. Moreover, MTTF component may affect the operation of another. For example,
represents the time when 63.2% of the items (under constant the reduction of capacitance will increase the associated
failure rate condition) would fail and varies with operation voltage ripple, which may cause over voltage stress of
conditions and testing methods [3 8]. switching devices even though the capacitor itself can still
operate under a normal mode. Similarly, the deterioration of
input and output performance of a specific power electronic
converter may induce failures of other subsystems. Therefore,
Infant Mortality Wear Out it could be more difficult to determine the failure criteria of a
(Quality) (Durability) component or system in power electronics than in other
domains. The selection of the parameter as the failure
Useful Life
t
indicator and the corresponding criteria depends on specific
Random Failures, Constant Failure Rate design, operation condition and standard. For illustration, the
(Reliability) failure criteria for electrolytic capacitors can be set as 100%
increase of the equivalent-series-resistance (ESR) or 20%
Time
reduction of the capacitance. Different results could be
Fig. 3. Failure rates presented by bathtub curve during three distinct periods. obtained for different choices of the failure indicator.

34

Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on January 27,2025 at 12:53:26 UTC from IEEE Xplore. Restrictions apply.
 / ! \
/i\ Ideal case without to empirical failure analysis based on historical data, PoF
degradation

i/ approach requires the knowledge of deterministic science (i.e.

materials, physics and chemistry) and probabilistic variation
theory (i.e. statistics). The analysis involves the mission
profile of the component, type of failure mechanism and the
associated physical-statistical model. Table I gives examples
of wear out failure mechanisms for electronic components as
presented in [41].
D. Typical Distribution ofFailures and Source ofStresses in
Power Electronic Systems
Stress or strength To perform reliability-oriented design, it is worthwhile to
explore the major failure modes and failure mechanisms of all
Fig. 4. Load-strength analysis to explain overstress failure and wear out reliability-critical components. Fig. 5 (a) and Fig. 5 (b) show
failure.
the failure distribution among power electronic components
B. Load and Strength Analysis [42] and source of stresses that have significant impact on
A component fails when the applied load L exceeds the reliability [43]. It can be noted that capacitors and
design strength S. Load L here refers to a kind of stress (e.g. semiconductors are the most vulnerable power electronic
voltage, cyclic load, temperature, etc.) and strength S refers to components, which is also verified by the survey conducted
any resisting physical property (e.g. harness, melting point, in [21]. Temperature has the most significant impact on the
adhesion, etc.) [12]. Fig. 4 presents a typical load-strength reliability of power electronic components and systems.
interference evolving with time. For most power electronic Therefore, electrical-thermal analysis and simulation are
components, neither load nor strength are fixed, but allocated important and necessary to perform reliability-oriented
within a certain interval which can be presented by a specific design.
probability density function (e.g. normal distribution).
TABLE 1. FAILURE MECHANISMS, RELEVANT LOADS, AND MODELS
Moreover, the strength of a material or device could be TN ELECTRONICS.

Failure Relevant Failure

degraded with time. Theoretically, the probability of failure
mechanisms Failure sites
loads models
can be obtained by analyzing the overlap area between the

LlT, Tmeon,
load distribution and the strength distribution. Practically,
DTldt,
Die attach, wire Nonlinear
the exact distributions of load and strength are very often not Power law
bond ITAB, solder
available, Monte Carlo simulation as discussed in [12] can be Fatigue
f..V
leads, bond pads, dwell time, (Coffin-
applied to randomly select samples from each distribution and interfaces f..H, Manson)

f..v, T
compare them and thus roughly estimate the probability of Eyring
Corrosion Metallization M,
failure. (Howard)
Although the load and strength analysis may not ensure an Electromigration Metallization T.J Eyring (black)

f..V
accurate prediction of probability of failure due to the Conductive Between Power law
M,
uncertainty of their distributions, it provides insight into how filament formation Metallization (Rudra)

T
to reduce failure. Proactive measures can be taken in design Stress driven Eyring
Metal traces S,
diffusion voiding (Okabayashi)
phase by setting a reasonable design margin (i.e. selection of
VT
Time dependent Arrhenius
S) or managing load (i.e. active control of L during
dielectric Dielectric layers (Fowler-
operation). Therefore, degradation models and lifetime breakdown Nordheim)
T: temperature; H: humidity; f..: cyclic range; V: voltage; M: moisture; J:
models are necessary to estimate the failures at the end-of-life
in the initial design phase, which will be discussed later. current density; v: gradient; S: stress.
C. Physics-oi-Failure Approach
A paradigm shift in reliability research on power
electronics is going on from today's handbook based methods
to more physics based approaches, which could provide better
understanding of the failure causes and design deficiencies,
so as to find solutions to improve the reliability rather than
obtaining analytical numbers only. Physics-of-Failure (PoF)
approach is a methodology based on root-cause failure
mechanism analysis and the impact of materials, defects and
stresses on product reliability [40]. Failure mechanisms can
be generally classified into overstress and wear out.
(a) Failure root cause distribution. (b) Source of stresses distribution.
Overstress failure arises as a result of a single load (e.g. over
voltage) while wear out failure arises because of cumulative Fig. 5. Failure and stress distributions in power electronic systems (Data
source: (a) from [42] and (b) from [43]).
damage related to load (e.g. temperature cycling). Compared

35

Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on January 27,2025 at 12:53:26 UTC from IEEE Xplore. Restrictions apply.
 Design

-- -- - ------------ - -- - --

:� �� �� 1 :� ��::�:::��:", �:
·
'

••
s
• tr
• Gi
� � :
. It "
electrical, thermal, mechanical)
(

:
•

�I.
Failure Mode and Effect
Analysis (FMEA)
• Fault tolerant design

::
• Load management design

; /
Critical component list and Virtual prototype

"' � '"" f
::
• •

Concept critical failure mechanisms • Initial reliability and durability

Key manufacture process analysis
��:���; � : �� � �,,
•
• Mission profile
" de O b 'I" ' "Iy "'
Topology and system
---
•
._ ___ _ _ -- - - - -----------
architecture _ -,
• Risk assessment
(e.g. new technology, , __________________L __________________ \ .. , .. ... ______________________________________.... ,
new components)
,
Verification i Optimized Design

:::
• Existing database ..

Prototype building Life cycle cost

::
• •

• Accelerated Lifetime Testing • Reliability

',i::: . .
(ALT) • Durability (i.e. lifetime)
Multi Environment Over Availability
:
•

Stress Testing (MEOST) • Robustness

• Testing data analysis • Power efficiency
• Degradation analysis • Power density

l
• Reliability and durability • Control performance
, analysis
..' _
.... ------------------------------------- _..

.. I
+
.. _-----------------------------------_

Validation Production Release

• System level • Process control • Customer usage

functionality testing • Process FMEA • Condition monitoring
• ALT • Screening testing • Field data
• Highly Accelerated (e.g. Highly • Root cause analysis
Limit Testing (HALT) Accelerated Stress data
• Robustness Screening (HASS)) • Corrective action
validation data

I
I I

L _________________ �____________________________________________________ a
I I

Fig. 6. Proposed design for reliability procedure for power electronic systems.

III. PROPOSED DFR PROCEDURE FOR POWER ELECTRONIC profile based on record data of a specific location and time
SYSTEMS AND AS SOCIATED DESIGN TOOLS period. Fig. 7 gives an example of the wind speed distribution
A systematic DFR procedure specifically applicable to
at the Lee Ranch facility in Colorado during the year of 2002

---- ---- J:------ frequency

[44].
power electronic system design is proposed as given in Fig. 6.
It can be noted that the procedure designs reliability into each
800,- ---- ----
� -� � � --,

development process (i.e. concept, design, validation, 700 energy

production and release) of power electronic products, 600
especially in the design phase. Therefore, attention is given
to the detailed procedures and various design tools applied in '-�
If)
500

400
the design phase according to the initial design concept. ..c
300
A. Concept Phase

�
200
In the initial concept phase, the relevant conditions to 100
which the power electronic systems are expected to be m
0
exposed (i.e. mission profile) are identified. Benchmarking 5
wind speed (m/s)
0 10 15 20 25
of system architecture and circuit topology is conducted.
Then, the potential new risks brought into the design are
Fig. 7. Distribution of wind speed (red) and energy generated (blue) for all
analyzed based on past experience and applied new type of of 2002 at the Lee Ranch facility in Colorado. The histogram shows
devices and topologies. In the wind power application, the measured data, while the curve is the Raleigh model distribution for the
mission profile mainly depends on the wind speed profile. same average wind speed (source: Wikipedia adapted from [44]).
Therefore, it is necessary to analyze the feature of the wind

36

Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on January 27,2025 at 12:53:26 UTC from IEEE Xplore. Restrictions apply.
 ToolBox

(-------------c-��bi��-d-iVi�d�i�-------------·l
l. Statistical model (e.g. Weibull distribution) l
i. Physical models i
i i
Input I . Cumulative damage model

•
Mission profile
Stress and strength
Critical component
list and critical failure
mechanisms
;:: : : : : ::��;�;�;:�:�;;�; �;:: : : �
i
l.
. Available data analysis
Parameterization method (e.g. Maximum
i
l L....o...
•
Output

Reliability (failure
rate)
t .: ,-----
.��--_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ f_ -_--_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ -_ :: .
likelihood method) • Durability (lifetime)
• Robustness
(robustness

r Component level
1
indication factor)
Input II
i. Reliability prediction for each critical i
• Simulation test data .: failure mechanism .:

l� �::��:���:��:�:::::::::t:::::::::::::::::::::::::::i
• Acclerated testing :. Reliability prediction for each critical :
data
• Field data

: System level :
: :
l· System reliability model
l
_
...... --------------------------------------------------�'i
Analysis methods (e.g. block diagram
i.
• • •

analysis, fault tree analysis)

Fig. 8. Proposed reliability prediction procedure for power electronic systems.

B. Design Phase - Analysis

The analysis covers the following four aspects: a) basic then the whole system. Statistical models are well presented
operation of the power electronic circuit and system; b) in [12] and will not be discussed here. Temperature and
electrical and thermal stress analysis based on the system temperature cycling are the major stressors that affect the
specifications and mission profile for preliminary selection of reliability performance as shown in Fig. 5, which will be
components to meet the stress-strength requirement; c) more significant with the trend for high power density and
Failure Mode Effect and Analysis (FMEA) [12] to identify high temperature power electronic systems. Therefore, two
the failure mechanisms as shown in Table I. failure mode models are presented here to study their effects.
(e.g. open circuit. short circuit. etc.). occurrence and severity a) Degradation model on the temperature effect
level of the failure and likelihood of prior detection for each
Fig. 9(a) illustrates the degradation of a material or device
cause of failure and d) list of reliability critical components in
from initial stable state with free energy of E\ to a degraded
the system and their associated failure mechanisms,
state with free energy of E2, The driving force for this
C. Design Phase -Initial Design degradation is the free energy difference between E\ and E20
Multi-domain simulation, especially the electrical-thermal defined as /I.E, The heat induced by power losses in power
simulation is a very useful tool to virtually investigate the electronic components provides the energy for the
static and dynamic properties of the system to be designed as transformation from one state to another. The rate of the
discussed in [45]-[46]. The link between the electrical domain degradation is limited by the activation energy Ea. Define
and thermal domain is the power loss and thermal model of kji!ll'ard, kreverse and knef as the degradation rate, recovery rate
individual component. Finite Element Analysis (FEA) can be and net reaction rate, respectively. It can be derived that
used to study the thermal distributions,
Fault tolerant design is a way to reduce the system level
failures for some critical applications (e,g, data center) (1)
requiring high level of reliability, Due to the redundancy
design as summarized in [3 0], a fault in a component or
subsystem does not induce the failure of the whole system, whereEa =-KB
KB is the Boltzmann's [aln(knel)/a(lI T)] ,
therefore, preventing the system from significant loss or s
constant (8.62xl0- eV/K), Tis the temperature in Kelvin and
unexpected interruptions. ko is a material/device specific constant. It should be noted
At this stage, an initial reliability prediction can be that the simplified result in (1) is the same as Arrhenius
performed. Fig. 8 proposes a generic prediction procedure equation that is widely used for reliability prediction, The
based on the PoF approach. The toolbox includes combined value of the activation energy is dependent on the type of
models and various sources of available data (e.g. material and device.
manufacturer testing data, simulation data and field data, etc.)
for the reliability prediction of individual components and

37

Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on January 27,2025 at 12:53:26 UTC from IEEE Xplore. Restrictions apply.
 Bond Bondwire

/ /0"
Forward reaction IGBT
)
Diode
Reverse reaction
(
Chip Solder
f
; �} Substrate
�];;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;�
------

>.
Base Plate SOlder
�
IE
Q)
Thermal Grease''::=::::J,;:
c B ase pla te
I a
Q)
Q)
==========::== =====:::JL

I
�
LL
Heatsink
-l
I f,.E
--� Fig. 10. Structural details of an IGBT module (connections that are relevant
Degraded state to module lifetime are marked red) [47].

Reaction coordinates ':.-- ElasticBehavior

,
(a) Free energy description of material/device degradation.

Forward reaction CSyield

Reverse reaction PlasticBehavior

-r
>-
e> e
Q)
-----

c
Q) I
Q) I Ea=Eao-a�
�
lJ... I

-t
I
t£ =t>Eo+b�
I
j

Reaction coordinates Fig. 1 1. A typical stress-strain (a-E) curve for a material [48].
(b) The impact of additional stress on material/device degradation.
Thermal cycling is found to be one of the main drivers for
Fig. 9. Material/device degradation from free energy perspective (adapted
failure of IGBT modules. The effect of the temperature
from [48]).
cycling can be explained by the typical stress-strain curve in
When combined stresses are applied, the activation energy Fig. II. (J is defined as the cyclic stress (e.g. temperature
M is dependent on additional applied stress � (e.g. electrical cycling) and E: is defined as the deformation. With a low
stress, mechanical stress, and chemical stress) as shown in cyclic stress below (Jyield, no damage occurs and the material is
Fig. 9(b). The parameters a and b are determined from stress­ in the elastic region. When the stress is increased above (Jyield,
induced degradation testing data. a is temperature dependent an irreversible deformation is induced and the material enters
a o + alKBT. It can be obtained that into the plastic region. The coefficients of thermal expansion

( J
and defined as a =

for different materials in the IGBT modules are different,

knet. -
- ko -- �e
b -[:�J� and under leading to stress formation in the packaging. The degradation
KRT will continue with each cycle until the material fails. The
low � and high �, respectively. Therefore, a power law number of cycle to failure for temperature cycling can be
dependence for stress � is used to bridge the gap between low obtained as
stress (knet is linear with �) and high stress (knet is exponential
with �). That is (3)

where k and m are empirically-determined constants and N is

(2) the number of cycles to failure. fl.T is the temperature cycle
range and fl.To is the portion of fl.T that in the elastic strain
b) Lifetime model on the temperature cycling effect range. If fl.To is negligible compared to fl.T, it can be dropped
The thermal cycling is a response to the converter line and out from the above equation and the equation turns to be the
loading variations as well as periodically commutation of Coffin-Manson model as discussed in [15].
power switching devices. It will induce cyclic temperature Constant parameters in the combined models can be
stress on different layers of materials used for fabrication of estimated according to the available data. Therefore, the
power electronic components. For example, Fig. 10 shows reliability of each critical individual component is predicted
the typical structure details of IGBT modules. by considering each of its associated critical failure

38

Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on January 27,2025 at 12:53:26 UTC from IEEE Xplore. Restrictions apply.
mechanism. To map the component level reliability TABLE II. CONVERTER PARAMETERS FOR CASE STUDY.
prediction to the system level, the system modeling method Topology 2L-BTB as shown in Fig. 12 (a)
reliability block diagram(RBD), fault-tree analysis (FTA) or
Rated output active power Po 2.3 MW
state-space analysis (e.g. Markov analysis) is applied as
discussed in detail in [3 0]. DC bus voltage Vd, I.I kV DC
'
Rated primary side voltage Vp 690V rms
IV. CASE STUDY ON A 2.3 MW WIND POWER CONVERTER Rated load current Itoad 1.93 kA rms
Switching frequency.fc 1 950 Hz
A. Topologies for Wind Power Converter
Filter inductance Lf 132/lH
As the state-of-the-art and most adopted solution for wind
power generation, single-cell partial-scale power converter is 1.6 kAII. 1 kVII25°C, 2 in
IGBT Selection I (grid side)
parallel
used in conjunction with the Doubly-Fed Induction Generator
(DFIG) [2]. Another configuration that is becoming popular 2.4 kA ll.lkVI 150aC, single
IGBT Selection 1/ (grid side)
switch
in the wind power application is a single-cell full-scale power
converter with asynchronous generator, electrically excited * Line-to-line voltage in the primary windings of transformer

synchronous generator (WRSG) or permanent magnet excited 11

type (PMSG) [2].
Three-phase converters are dominant for wind power 10

application to handle high power and reduce the energy of cost 9

$'
compared to single phase ones. Pulse width modulation­
8
.s
voltage source converter with two-level output voltage (2L­
PWM-VSC) is the most frequently used three-phase power '0 7

�
Q)
Q)
converter topology in wind power systems. As the interface 6
between the generator and power grid, two 2L-PWM-VSCs '0

§ 5
c:
are usually configured as a back-to-back structure (2L-BTB)
with a transformer on the grid side, as shown in Fig. 12 (a). A 4
technical advantage of the 2L-BTB solution is the relatively
3
simple structure and few components, which contributes to a
well-proven robust and reliable performance. 2
Three-level neutral point clamped back-to-back converter o 200 400 600 800 1000 1200 1400 1600 1800

(3L-NPC-BTB) topology is one of the most commercialized Time (second)

multi-level converter topologies on the market, which is (a) Wind speed profile for the case study.
shown in Fig. 12 (b). It achieves smaller size of filters and 2500
higher voltage handling capability. However, it is found that
the loss distribution is unequal between the outer and inner
52000 /'
/
switching devices in a switching arm, which might lead to ..><:

1500
uneven lifetime of power switching devices [49]- [51]. ....
OJ
3:

/
o
0..

� 1000
'0

/
co
....
OJ

Jl 500
c

Wind turbine
/
5 15 20 25
o
o 10 30
Wind speed (m/s)
(a) Two-level back-to-back converter (2L-BTB). (b) Power curve against wind speed of the chosen 2.3 MW wind turbine.
Fig. 13. Wind speed profile and power curve of the wind turbine.

B.

��'
Case Study
a) Wind speed profile and converter specifications
Wind turbine
Table II shows the parameters of a 2. 3 MW wind power
converter for the case study. For illustration purpose, the half
hour wind speed profile shown in Fig. 13(a) is analyzed. It
(b) Three-level neutral point clamped back-to-back converter (3L -BTB).
should be noted that for a specific practical application, wind
Fig. 12. State-of-the-art circuit topologies for wind power converters [2]. speed patterns are obtained by collecting wind speed profile

39

Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on January 27,2025 at 12:53:26 UTC from IEEE Xplore. Restrictions apply.
during a much longer time period. Fig. 13(b) shows the 109 Fr=====�U"�'�"fi�,".Tl===i-r.M�OO�"==�U"�"�'fi�"�2==n
Model

o-
Equation k"(x-xO)'i-n)

power curve against wind speed for the chosen wind turbine. Equation k"x'·n

4,40a14El1
Reduced
Chi.Sq.
3.21397E8

Reduced

IGBT modules as the reliability-critical components will be t l'ld rd

Chi·Sqr Adi.R-Square 0.99999

1-
Adj. R-Squ8re 0.9924 Value Sa a Err r
108
focused in the case study. Value
1.39127E16
Stal'ldardError
3.07122E16
,0

k
24.9934
4.9849E8
0.2226

8,99532E7

�
6 0.68072 0.0548

b) Lifetime prediction models

I I
The reliability of the IGBT modules in the grid side
converter is investigated. There are three dominant failure
mechanisms for IGBT modules when temperature cycling is
applied: baseplate solder joints cracking, chip solder joint
U
a
�
'0
>-

1l
10'

10· -
r�
�,
Sx 108 X (t.T.OGBT � 2S r' 02
' '-,

= 1.4
l - ---l
x I o -, 6 X--t. T. -

I
-
6
-1
,GB7. -+--+

J_

E N
�I
��
=

:::l
cracking and the wire bonds lift-off [52]. Based on the PoF
approach, the specific lifetime model is required for each
Z
(for t.T.OGBT > 25)
I
Testing dat, a at �ma,= 190 ° C and tpn = to, �60s
.

105 �----------+-------+�--�----_� 1 --+�,�r--+-1

failure mechanism. Therefore, three models are to be •

calibrated from (3), which considers both the plastic and - - - Fitting curve based on COffin-Ma
l nson m .o
1 del
-- Fitting curve based on (3 )
elastic behavior of the associated materials. 10 4 ������������ ���������iliw�
Moreover, as the amplitude and average temperature level 20 30 40 50 60 70 80 90 100

of the thermal cycling are different when the wind speed is

=
fluctuating, the Palmgren - Miner linear cumulative damage
120s and '0m== 100°C.
model [53] is applied in the form of (b) Cycle to failure models for thermal cycling on chip solder joints for t'yde

n
109 User define 1 Userde�ne2

x"
MOO"

u
Model

(4) Eq alio k· -n
Equalion

Reduced
k-(x·xO)",(-n)
449375.05901
Reduced 1,47751E12 Chi-Sq,
Chi_Sq.
Adj. R -S quare
Adj. R-Square
,
0.97544
Value StandardError
10' Value Stal'ldardErrOf
-
'0 33.20653 0.00854
-

k 4E17 1,90409E18
where nj is the number of applied temperature cycles at stress 0
6.48024 1.36404
0 k 8.5237E8
1.99913
4.88605E8
0.00165

111; and Ni is the number of cycles to failure at the same stress I

I d [[ [
z
and for the same cycle type. Therefore, each type of 111; ';;; 107 ,"" -, , -6.48
4 X 10 17 ilTL'GBT
, ---
Q) " N = X
'0
accounts for a portion of damage. Failure occurs when the >­
u
sum of the left hand side of (4) reaches 1. '0
Qj 10·
33.2 r'
( il TO �
-

c) Parameter estimation o/lifetime models .0 8

E N = 8.5 X lOx GBT
0,
(for 33.2)
�

1"
:::l
Z

I
The manufacturer of the IGBT modules has performed a ilTUGBT >
,

7Jmax = lOOOC and

mr
series of power cycling experiments as described in [52]. 105 , , ,
• Testing dat� at _ I-
ton = toft =6
Based on the B 10 lifetime testing data (which is the number of ,

T
- - - Fitting curve based on COl -Manson el
cycles during which 10% of the total number of modules - Fitting curvi based on (3)

fail), the parameters of the derived model in (3) can be 10'

30 40 50 60 70 80 90 100

LJTUGB-r<°C)
estimated as shown in Fig. 14.
10'
Model

Equat on

Rod_
Chi·Sq,
i x
k· " · n
User define 1

3.Z3382El0
Model
Equation

Reduced
Chi-Sq,

Adj. R-Square
User define 2
k"(x-xO)"(-n)

2.15381E6

1
(c) Cycle to failure models for thermal cycling on wire bonds T;max = 100°e.
Fig. 14. Parameter estimation based on the model derived in (3) and the
Adj.R-Square 0 99093 Value StandardEfTOf
Value Standard Ero
r r " 15 82623 0.07037 Cotlin-Mansion model (lifetime testing data source: [52]).
10' 1 0
k 3E12 4.44061E12 k a.S39NE7 5.47075E6.

�
,
0
4.43902 0S0652 _ 1.9951 0.0193

1 T T 1
The lifetime testing data are also used to fit the widely

�'"�1 " "

If) N 8 . 8x I 0' X (t.7; I S .8 f'
Q)
= �
used Coffm-Mansion model, in which the yield amplitude of
'0
>- (fort.7; > 15.8)
u temperature cycle, thus, the elastic behavior is neglected.
10·
a N 3x 10" Xt.7;-4·44 However, with high number of cycling (> 104), thus, with low
Qj
=

�" I
.0 cyclic temperature stress, the absence of the elastic behavior
E
Z
::J
may induce high level of inaccuracy. As shown in Fig. 14,

L, �1'::-::---
105 there is a considerable discrepancy between the two models,
• "" " "" '" �"i f C �
of which the model in (3) can fit the testing data well. The
estimated value of I1To in (3) implies that it has negligible
cmm

I
I 1 on 0 1 ....
- - - Fitting curve based on cot -Manson model '

T
.
'

(3) ' I
'
- Fitting curve based on I ,
impact on the lifetime for temperature cycling with amplitude
10'
20 30 40 50 60 70 80 90 100 less than 11To .
Moreover, the number of cycles to failure is also
dependent on the average junction temperature '0 and average
tcecle = 120s and Temin = 60°e.
(a) Cycle to failure models for thermal cycling on baseplate solder joints for
case temperature of Te. A scaling factor is provided by the
manufacturer in [54] as

(/1J116)
Scalmg factor 1.017
•

=
,,,,' (5)

40

Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on January 27,2025 at 12:53:26 UTC from IEEE Xplore. Restrictions apply.
where !'!,.Tzeve/ is the difference between two maximum junction
temperature levels or two minimum case temperature levels.
d) Distribution of Temperature profile and temperature
cycling 80

� 60
<II
Fig. 15 (a) and (b) show the case temperature and junction '"

temperature for IGBT modules of Selection 1 and Selection 11, ()

40
respectively. It can be noted that IGBT modules with a lower '0
W
current rating have higher amplitudes of !'!,.T; and !'!,.Tc, thus,
20
.c
E
::l
lower number of cycles to failure. Z

To perform the lifetime prediction, the analysis of the

o
temperature cycling distribution is necessary. The Rainflow o
counting method [55] is applied to extract the temperature
information as shown in Fig. 16. Fig. 16(a) and (b) show the
case temperature and junction temperature cycling with the
IGBT modules of Selection 1. Fig.16(c) and (d) give those of
Mean TjlGBT
the IGBT modules of Section 11. It can be noted that most of
the temperature cycling is of very low amplitude which has
kAII.IkV IGBTs in parallel (Temperature unit: 0 c) .
(a) Distribution ofjunction temperature cycling with Selection loftwo 1.6
negligible impact on the lifetime.
120
- . . . ;. ..... .
110

100
100

� 80
'"
90 <II
'0

60
� >-

�..
::l ()
80
'0
40
Q. Ci;
�
E 70 .0

20
E
:::l
60 z

0
50

40
0 200 400 600 800 1000 1200 1400 1600 1800
Time (second)
(a) The IGBT junction temperature and case temperature with Selection lof
two 1.6 kAII. 1 kV IGBTs in parallel.

kAll.lkV IGBTs in parallel (Temperature unit: 0 c) .

(b) Distribution ofcase temperature cycling with Selection I oftwo 1.6
120

110

100

E 90
�
E � 60
80 '0
� >-

40
..
Q. ()

�
E 70 '0
Q)
§ 20
.0
60
z
50
o
40
o 200 400 600 800 1000 1200 1400 1600 1800
Time (second)
(b) The IGBT junction temperature and case temperature with Selection II of
single 2.4 kA 1I.lkV IGBT.

Fig. 15. Temperature profile of the two selected IGBT modules.

kA ll.lkV IGBT (Temperature unit: 0 c) .

(c) Distribution ofjunction temperature cycling with Selection II ofsingle 2.4

41

Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on January 27,2025 at 12:53:26 UTC from IEEE Xplore. Restrictions apply.
 C. Methods to Improve Reliability
a) Selection ofproper devices
The results shown in Table III imply that the selection of
IGST modules has significant impact on the lifetime.

l(l
Comprehensive analysis on the device selection based on
60 both cost and performance is needed to avoid either over
u
>-
u engineering design or fail to meet the specifications.
'0 40
Q; b) Low Voltage Ride Through (L VRT) thermal optimized
.n
§ 20 modulation
z
Thermal loading of the power device can be improved by
o
the modulation schemes. Some modulation methods for
50 thermal optimization of 3L-NPC-BTB wind power converters
during extreme LVRT are proposed in [56]. The basic idea of
these modulations is to select the proper vector sequences
which can reduce the dwelling time or commutations
40 involving zero voltage level. The loss and thermal in the
most stressed devices can thereby be reduced.
(d) Distribution ofcase temperature cycling with Selection II ofsingle 2.4 kA
1 1 . 1 kV IGBT (Temperature unit: DC). c) Reactive power control during wind gust
Fig. 1 6. Rainflow counting ofthe temperature profiles for IGBT modules. The amplitudes of !'!.Tc and !'!.Tj have significant impact on
the lifetime of IGBT modules. To limit the increase of the
e) Lifetime prediction results thermal cycling stress during wind gust, the possible ways to
According to the calibrated models shown in Fig. 14, only control the reactive power and reduce the thermal loading is
the temperature cycling with !'!.Tc > 15. 8°C, !'!.T; TUBT > 25°C discussed in [57]. The normal operation mode and reactive
and !'!. T; TUBT > 3 3.2°C are considered for the lifetime power control mode of the converter is switched according to
prediction for baseplate solder joints, IGBT chip solder joints the grid condition.
and wire bonds, respectively. The impact of the ones with d) On-line condition monitoring
lower amplitude is sufficiently to be neglected. For Condition monitoring is an effective way to enhance
illustration purpose, the wind profile is assumed to repeat as reliability when the power converters are in operation [58].
that in the studied 3 0 minutes and the wind turbine operates It provides the real-time operating characteristics and health
24 hours per day and 365 days per year. According to the conditions of the systems by monitoring specific parameters
models estimated in Fig. 14 and the ones shown in (4) and of power electronic components (e.g. saturation voltage of
(5), the lifetime of the two selected IGST modules is given in IGSTs). Therefore, proactive maintenance work could be
Table III. planned to avoid failures that would occur.
It should be noted that the purpose of the study case on
the IGST modules is to demonstrate the procedure to perform CONCLUSIONS
reliability prediction based on mission profile and PoF More and more efforts have been devoted to better power
approach with differentiation of various failure mechanisms. electronic systems in terms of reliability to ensure higher
For practical considerations, wind speed profile during long availability, more power generation and low maintenance
time period at specific location should be analyzed. cost. A paradigm shift in reliability research on power
Therefore, wind profile as shown in Fig. 7 could be useful for electronics is going on from simple handbook based
future research. Moreover, other failure mechanisms in IGBT calculations to the physics-of-failure approach and design for
modules may induce additional failures and should also be reliability process. A systematic design procedure consisting
considered in the lifetime estimation. of various design tools is presented in this paper to give an
outline on how to design reliability into the power electronic
TABLE III. LIFETIME PREDICTION RESULTS.
products from the early concept phase. The case study on a
lifetime (year)
Failure mechanisms
B10 2 . 3 MW wind power converter demonstrates some aspect of
Selection I Section II
the design procedure with emphasis on the lifetime prediction
I'1T, )
of the two kinds of iGST modules. It is based on analysis on
Baseplate solder joints (due to 358 24
1'1
IGBT chip solder joints (due to '0 1GJJ1)
mission profile, failure mechanism, thermal profile and
438 22
1'1 '0 1GB] )
estimation of the associated lifetime models.
Wire bonds (due to 2633 74 The major challenges and opportunities in the research on
Overall (determined by the shortest one) 358 22 reliability of power electronic systems are addressed.
A. Challenges
a) Outdated paradigms and lack of understanding in
design for reliability approach.

42

Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on January 27,2025 at 12:53:26 UTC from IEEE Xplore. Restrictions apply.
 b) Uncertainties in mISSIOn profile and strength of [ 1 1 ] L. M. Moore and H. N. Post, "Five years of operating
components. experience at a large, utility-scale photovoltaic generating
plant," Journal of Prog. Photovolt: Res. App l. , vol. 1 6, no. 3 ,
c) Increasing electrical/electronic content and complexity.
p p . 249-259, 2008.
d) Lacking of understanding of failure mechanisms and
[ 1 2] P. O'Connor, A. Kleyner, Practical reliability engineering, the
failure modes of reliability critical components. h
5t edition, John Wiley & Sons, 20 1 2, ISBN: 9780470979822.
e) Resource-consuming testing for reliability prediction
[ 1 3 ] L. G. Reddy, L. M. Tolbert, B . Ozpineci, X. Yan, and D . T.
and robustness validation.
Rizy, "Reliability of IGBT in a STATCOM for harmonic
B. Opportunities compensation and power factor correction," in Proc. IEEE
App l. Power Electrons. Con! and Exp o. , pp. 783-788, 20 1 2 .
a) Better mission profile and on-line monitoring data
[ 1 4] V. Smet, F . Forest, J. Huselstein, F . Richardeau, Z. Khatir, S .
from the field.
Lefebvre, and M. Berkani, "Ageing and failure modes of IGBT
b) Physics-of-failure approach provides insight into how
modules in high-temperature power cycling," IEEE Trans. on
to avoid failures in power electronic systems. Ind. Electrons. , vol. 58, no. 1 0, pp. 493 1 -494 1 , Oct. 20 1 1 .
c) Control of heat flow and thermal distribution by
[ 1 5] c . Busca, R. Teodorescu, F . B laabj erg, S . Munk-Nielsen, L.
controlling the power in power electronic circuits. Helle, T. Abeyasekera and P. Rodriguez, "An overview of the
d) Component level and system level smart de-rating reliability prediction related aspects of high power IGBTs in
operation. wind power applications," Journal of Microelectronics
e) Condition monitoring and fault tolerant design allow Reliability, vol. 5 1 , no. 9- 1 1 , pp. 1 903 - 1 907, 2 0 1 1 .
extended lifetime. [ 1 6] A. Testa, S . D e Caro, and S . Russo, "A reliability model for
f) Emerging semiconductor and capacitor technologies power MOSFETs working in avalanche mode based on an
ensure more reliable power electronic components. experimental temperature distribution analysis," IEEE Trans.

g) Computer-aided automated design software to save on Power Electrons. , vol. 27, no.6, pp . 3 093-3 1 00, Jun. 2 0 1 2 .

time and cost. [ 1 7] J. Wuertl, E. B ahat-Treidel, F . Brunner, E. Cho, O . Hilt, P . Ivo,

A. Knauer, P . Kurpas, R. Lossy, M. Schulz, S . Singwald, M.
REFERENCES Weyers, and R. Zhytnytska, "Reliability issues of GaN based
high voltage power devices," Journal of Microelectronics
[I] F. Blaabj erg, Z. Chen, S. B. Kj aer, "Power electronics as
Reliability, vol. 5 1 , no. 9 - 1 1 , pp. 1 7 1 0- 1 7 1 6, 2 0 1 1 .
efficient interface in dispersed power generation systems,"
IEEE Trans. on Power Electrons. , vol. 1 9, no. 4, pp. 1 1 84- [ 1 8] M . A. Vogelsberger, T . Wiesinger, and H . Ertl, "Life-cycle
1 1 94, Sep. 2004. monitoring and voltage-managing unit for dc-link electrolytic
capacitors in PWM converters," IEEE Trans. on Power
[2] F . Blaabj erg, M. Liserre, K. Ma, "Power electronics converters
Electrons. , vol. 26, no. 2, pp. 493 - 5 0 3 , Feb. 20 1 1 .
for wind turbine systems," IEEE Trans. on Ind. App l. , vo1.48,
no.2, pp.708-7 1 9, Mar-Apr. 20 1 2 . [ 1 9] A. Lahyani, P . Venet, G. Grellet, and P . J. Viverge, "Failure
prediction of electrolytic capacitors during operation of a
[3] European Commission Climate Action, "The E U climate and
switch mode power supply," IEEE Trans. on Power Electrons. ,
energy package," Mar. 2007.
vol. 1 3 , no. 6, pp. 1 1 99- 1 207, Nov. 1 99 8 .
[4] S . Faulstich, P . Lyding, B . Hahn, P . J. Tavner "Reliability of
[20] G. M. Buiatti, S . M. A. Cruz, and A. J. M. Cardoso, "Lifetime
offshore turbines- identifying the risk by onshore experience,"
of film capacitors in single-phase regenerative induction motor
in Proc. of Europ ean Offshore Wind Energy Conference and
drives," in Proc. IEEE International Symp osium on Diagnostics
Exp osition, Stockholm, 2009.
for Electric Machines, Power Electronics and Drives, pp. 3 56-
[5] B. Hahn, M. Durstewitz, K. Rohrig "Reliability of wind 3 62, 2007.
turbines - Experience of 1 5 years with 1 5 00 WTs," Wind
[2 1 ] S . Yang, A. T. Bryant, P . A. Mawby, D . Xiang, L. Ran, and P.
Energy: Proceedings of the Euromech Colloquium, pp. 3 29-
Tavner, "An industry-based survey of reliability in power
3 3 2, Springer-Verlag, Berlin.
electronic converters," IEEE Trans. on Ind. App l. , vol. 47, no.
[6] R. Johan and B. L. Margareta, "Survey of failures in wind 3, pp. 1 44 1 - 1 45 1 , May/Jun. ,20 1 1 .
power systems with focus on Swedish wind power plants
[22] D . Hirschmann, D . Tissen, S . Schroder, and R. W . D e Doncker,
during 1 997-2005," IEEE Trans. on Energy Convers. , vol. 22,
"Reliability prediction for inverters in hybrid electrical
no. I , pp. 1 67- 1 73 , Mar. 2007.
vehicles," IEEE Trans. on Power Electron. , vol. 22, no. 6, pp.
[7] P. J. Tavner, J. Xiang and F . Spinato, "Reliability analysis for
25 1 1 -25 1 7, Nov. 2007.
wind turbines," Wind Energy, vol. 1 0, no. l , pp. 1 - 1 8, 2007.
[23] M. Boettcher and F . W. Fuchs, "Power electronic converters in
[8] S . B . Kj aer, J. K. Pedersen and F . B laabj erg, "A review of
wind energy systems -Considerations of reliability and
single-phase grid connected inverters for photovoltaic strategies for increasing availability," in Proc. European
modules," IEEE Trans. on Ind. App l. , vol. 4 1 , no. 5, pp. 1 292-
Conference on Power Electronics and App lications, pp. 1 - 1 0,
1 3 06, Sep. 2005. 20 1 1 .
[9] M. Schmela, "PHOTON : Inverter survey 20 1 2 stats," Presented
[24] R. Luczkowski and R. Muszynski, "Cost and reliability
at the PHOTON 's 3rd PV Inverter Conference, San Francisco, oriented design of the converter for small wind power plant," in
Feb . , 20 1 2 . Proc. lET Colloquium on Reliability in Electromagnetic
[ 1 0] T. McMahon, G . Jorgensen, R . Hulstrom, "Module 3 0 year life: Systems, pp. 1 -5 , 2007.
what does it mean and is it predictable/achievable?" National [25] E. Koutroulis and F. Blaabj erg, "Design optimization of
Renewable Energy Laboratory, Reliability Physics Symp osium, transformer-less grid-connected PV inverters including
pp. I 72- 1 77, 2008. reliability," IEEE Trans. on Power Electron. , forthcoming and

43

Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on January 27,2025 at 12:53:26 UTC from IEEE Xplore. Restrictions apply.
 online available http://ieeexplore. ieee. org. [4 1 ] B. Tuchband, N. Vichare and M. Pecht, "A method for

[26] S. V. Dhople, A. Davoudi, A. D. Dominguez-Garci and P. L. implementing prognostics to legacy systems," in Proc. IMAPS

Chapman, "A unitied approach to reliability assessment of Military, Aerospace, Space and Homeland Security: Packaging
multiphase dc-dc converters in photovoltaic energy conversion Issues and App lications, 2006.

systems," IEEE Trans. on Power Electron. , vol. 27, no. 2, pp. [42] E. Wolfgang, "Examples for failures in power electronics
739- 75 1 . Feb. 20 1 2 . systems," presented at ECPE Tutorial on Reliability of Power

[27] C . Rodriguez and G. A . J. Amaratunga, "Long-lifetime power Electronic Systems, Nuremberg, Germany, Apr. 2007.
inverter for photovoltaic AC modules," IEEE Trans. on Ind. [43] ZVEL, Handbook for robustness validation of automotive
Electron. , vol. 5 5 , no. 7, pp. 2593 -260 1 , Jul. 2008. electrical/electronic modules, Jun. 2008.
[28] F . Carastro, A. Castellazzi, J. Clare, and P . Wheeler, "High­ [44] Global Energy Concepts, " S andia national laboratories new
efficiency high-reliability pulsed power converters for Mexico wind resource assessment Lee Ranch," online available
industrial processes," IEEE Trans. on Power Electron. , vol. 27, http://windpower. sandia.gov/other/LeeRanchData-2002 .pdf,
no. I, pp. 3 7-45, Jan. 20 1 2 . Feb. 2003 .

[29] R . Burgos, C . Gang, F . Wang, D . Boroyevich, W. G. Odendaal [45] 1. Biela, 1. W. Kolar, A. Stupar, U. Drofenik, and A. Muesing,
and J. D. V. Wyk, "Reliability-oriented design of three-phase "Towards virtual prototyping and comprehensive multi­
power converters for aircraft applications," IEEE Trans. on obj ective optimization in power electronics," in Proc. of PCIM
Aerosp . Electron. Syst. , vol. 48, no. 2, pp. 1 249- 1 26 3 , Apr. 20 1 0, Nuremberg, pp. 1 -2 3 , 20 1 0 .
20 1 2 . [46] D . Bin, J. L. Hudgins, E. Santi, A . T. Bryant, P . R . Palmer, and
[30] Y. Song and B . Wang, "Survey o n reliability of power H. A. Mantooth, "Transient electrothermal simulation of power
electronic systems," IEEE Trans. on Power Electron. , semiconductor devices," IEEE Trans. on Power Electron. , vol.
forthcoming and online available http://ieeexplore. ieee. org. 25, no. 1, pp. 23 7-248, Jan. 20 1 0.

[3 1 ] G. Petrone, G. Spagnuolo, R. Teodorescu, M. Veerachary and [47] A. Wintrich, U. Nicolai, W. Tursky, and T. Reimann,
M. Vitelli, "Reliability issues in photovoltaic power processing App lication manual p ower semiconductors, SEMIKRON
systems," IEEE Trans. Ind. Electron. , vo1 . 5 5 , no.7, pp. 2569- International, pp. 1 22, 20 1 1 , ISBN: 9783 9 3 8 843666.
2580. Jul. 2008. [48] J. W. McPherson, Reliability physics and engineering: time-to­
[32] S . S . Smater and A. D . Dominguez-Garcia, "A framework for failure modeling, Springer, 20 1 0, ISBN: 978 1 44 1 963475 .
reliability and performance assessment of wind energy [49] R. Teichmann, S. Bernet, "A comparison of three-level
conversion systems," IEEE Trans. on Power Sys. , vol. 26, no.4, converters versus two-level converters for low-voltage drives,
pp. 223 5-2245, Nov. 2 0 1 1 . traction, and utility applications. " IEEE Trans. on Industry
[3 3 ] K . Fischer, F . Besnard, and L . Bertling, "Reliability-centered App lications, vol. 4 1 , no. 3 , pp . 855-865, 2005.
maintenance for wind turbines based on statistical analysis and [50] K. Ma, F. Blaabj erg, "Multilevel converters for 10 MW wind
practical experience," IEEE Trans. on Energy Conversion, vol. turbines," in Proc. ofEPE 20 I I , pp. 1 - 1 0, 20 I 1 .
27, no. I, pp. 1 84- 1 95, Mar. 20 1 2 . [5 1 ] T . Bruckner, S . Bernet, H . Guldner, "The active NPC converter

[34] K. Moslehi and R . Kumar, " A reliability perspective of the and its loss-balancing control," IEEE Trans. on Industrial

smart grid," IEEE Trans. on Smart Grid, vol. I , no. I , pp. 57- Electronics, vol. 52, no. 3, pp . 8 5 5-868, 200 5 .
™
64, Jun. 20 1 0 . [52] ABB Application Note, Load-cycling cap ability of HiPak
IGET modules, 20 1 2 .
[3 5] P. Wikstrom, L. A. Terens, and H. Kobi, "Reliability,
availability, and maintainability of high-power variable- speed [ 5 3 ] Miner, M . A . , "Cumulative damage i n fatigue, " Journal of

drive systems," IEEE Trans. on Ind. App /. , vol. 36, no. 1, pp. App lied Mechanics, no. 1 2, A I 5 9-A I 64, 1 94 5 .

23 1 -24 1 , Jan.lFeb. 2000. [ 5 4 ] ABB Application Note, Load-cycling capability of HiPaks,

[36] G. A. Klutke, Peter C . Kiessler, and M. A. Wortman. "A 2004.

critical look at the B athtub curve," IEEE Trans. on Reliability, [55] ASTM International, EI 049-85 (2005) Standard practices for

vol. 52, no. I , pp. 1 25 - 1 29, Mar., 2003 . cycle counting in fatigue analysis, 200 5 .

[37] J. Liu and N. Henze, "Reliability consideration of low-power [56] K. Ma, F. Blaabj erg, 'Thermal optimized modulation methods

grid-tied inverter for photovoltaic application," in Proc. of three-level neutral-poi nt-clamped inverter for 1 0 MW wind

Europ ean Photovoltaic Solar Energy Conference and turbines under low voltage ride through," lET Power
Exhibition, pp. 2 1 -25, 2009. Electronics, 20 1 2 (to appear).

[38] M. Krasich, "How to estimate and use MTTFIMTBF would the [57] K. Ma, M. Liserre, F. Blaabj erg, "Reactive power influence on

real MTBF please stand up?" in Proc. of IEEE Annual ReI. and the thermal cycling of multi-MW wind power inverter", in

Maintain. Symp . , pp . 3 5 3 - 3 5 9 , 2009. Proc. ofIEEE APEC 2012, pp. 262-269, 20 1 2 .

[39] Military Handbook: Reliability [ 5 8 ] S . Yang, D . Xiang, A . Bryant, P. Mawby, L. Ran and P .
prediction of electronic
equip ment, MIL-HDBK-2 1 7F, Dec. 2, 1 99 1 . Tavner, "Condition monitoring for device reliability in power
electronic converters: a review." IEEE Trans. on Power
[40] M . Pecht and A. Dasgupta, "Physics-of-failure : an approach to
Electron . , vol. 25, no. 1 1 , pp. 2734-2752, Nov . , 20 1 0 .
reliable product development," in Proc. International
Integrated Reliability Workshop , pp. 1 -4, 1 99 5 .

44

Authorized licensed use limited to: Universidade Estadual de Campinas. Downloaded on January 27,2025 at 12:53:26 UTC from IEEE Xplore. Restrictions apply.