This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2020.2981933, IEEE
Transactions on Power Electronics

A Guideline for Reliability Prediction in Power

Electronic Converters
Saeed Peyghami, Member, IEEE, Zhongxu Wang, Student Member, IEEE, and Frede Blaabjerg, Fellow, IEEE

Abstract—Reliability prediction in power electronic systems intensifies the importance of converter reliability
converters is of paramount importance for converter prediction [2], [8]. Furthermore, evaluating new converter
manufacturers and operators. Conventional approaches employ topologies/redundant operation [9]–[11], switching schemes,
generic data provided in handbooks for random chance failure and control algorithms [8], [10], [12], [13] as well as analyzing
probability prediction within useful lifetime. However, the wear- the impact of control and operating conditions on the long-
out failures affect the long-term performance of the converters.
Therefore, this paper proposes a comprehensive approach for
term performance of converters [2], [14] requires appropriate
estimating the converter reliability within useful lifetime and reliability models of the converter.
wear-out period. Moreover, this paper proposes a wear-out So far, different approaches have been used for converter
failure prediction approach based on a structural reliability reliability estimation [10], [12], [15]–[22]. The most common
concept. The proposed approach can quickly predict the used method relies on the Military Handbook 217 (MIL-
converter wear-out behavior unlike conventional Monte Carlo HDBK-217). The main concerns regarding MIL-HDBK-217
based techniques. Hence, it facilitates reliability modeling and are out-of-date data for new technologies, vagueness of the
evaluation in large-scale power electronic based power systems failure mechanisms and data type, and exclusion of different
with huge number of components. The proposed comprehensive operating conditions. Besides MIL-HDBK-217, some
failure function over the useful lifetime and wear-out phase can
be used for optimal design and manufacturing by identifying the
companies and organizations have updated this handbook data
failure prone components and end-of-life prediction. Moreover, and methodology, such as Telcordia SR-322, Siemens
the proposed reliability model can be used for optimal decision- SN29500, RDF-2000. All these approaches carry the MIL-
making in design, planning, operation and maintenance of HDBK-217 shortcomings even though they have some
modern power electronic based power systems. The proposed updates on this handbook. Later on, the International
methodology is exemplified for a photovoltaic inverter by Electrotechnical Commission (IEC) released IEC TR-62380
predicting its failure characteristics. [23], which considers the failure mechanisms for failure rate
Keywords— converter reliability, failure rate, wear-out failure, prediction throughout a mission profile. However, the
constant failure rate, reliability modeling, systematic failure, provided data are not still updated and the failure mechanisms
catastrophic failure. are not accurately modeled. Therefore, the IEC TR-62380 has
Introduction been replaced by IEC 61709 [24], which provides a general
guideline for mission profile based failure rate prediction.
Power electronics reliability has gained an increasing In the aforementioned handbook methods, the failure
interest recently due to the role it plays in the modernization mechanisms are not accurately modeled and physics of
of the future power grids [1], [2]. Power converters are the failures are not considered. Therefore, the predicted reliability
main energy conversion system in a wide range of applications may not be acceptable, and may not be suitable for reliable
such as renewable energies, energy storages, high/medium design of converter components. Moreover, identifying the
voltage Direct Current (DC) transmission systems, weak points for reliability enhancement is not clear. Hence,
medium/low voltage DC distribution systems and e-mobility another update on MIL-HDBK-217 has been provided by
[3], [4]. However, the converters seem to be the vulnerable FIDES group where the physics of failures are considered in
components according to industrial experiences [1], [5], [6]. the failure rate prediction [25]. So far, the FIDES approach is
Therefore, high proliferation of the converters will pose new the latest update on the failure rate prediction of electronic
challenges in terms of optimal and reliable design, planning, components.
operation, and maintenance of the future power grids. All the handbooks provide a constant failure rate for
An expected end-of-life of converters is of paramount components during their useful lifetime. It is assumed that the
importance for an optimal decision-making in planning of components are appropriately designed and they do not enter
modern power electronic systems [7]. The optimal facility the wear-out phase during the mission life period [23], [25].
planning including cost-effective design and replacement Moreover, the IEC TR-62380 has provided lifetime
scheduling depends on the converters lifetime. Moreover, the expectancy for the components prone to wear-out failure. In
converter failure rate will affect its availability and optimal spite of considering the mission profile in the IEC TR-62380
operational planning of power systems. The maintenance for constant failure rate prediction, it is not taken into account
scheduling for repair and replacement of power converters for the end-of-life prediction. Therefore, the life expectancy
requires appropriate reliability modeling. Moreover, reliability limits may not be accurate enough for different operating
modeling is an important task for designers to do optimal and conditions.
reliable converter manufacturing. As a result, the decision- On the other hand, wear-out failure analysis in converter
making on investment during manufacturing, system-level components based on physics of failures has been addressed
planning, operation and maintenance of power electronic recently in [17]–[21]. Particularly, the wear-out probability
prediction in converter components has been explored in [17],
This work was supported by the Reliable Power Electronic-Based Power [18]. A Monte Carlo Simulation (MCS) based technique is
System (REPEPS) project at the Department of Energy Technology, employed to model device aging. The employed method in
Aalborg University as a part of the Villum Investigator Program funded
[17], [18] relies on the MCS, where in practice the MCS
by the Villum Foundation.
suffers from computational burden. Specially, for large-scale
The authors are with the Department of Energy Technology, Aalborg power electronic based power systems, employing MCS for
University, Aalborg 9220 east, Denmark. (Email: [email protected], all the components in different converters with different
[email protected], [email protected]) mission profiles is almost infeasible. Moreover, on-line

0885-8993 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 27,2020 at 07:25:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2020.2981933, IEEE
Transactions on Power Electronics

reliability prediction for control purposes (e.g., in [8]) requires

a fast reliability prediction approach. Meanwhile, the
repeating MCS in design for reliability approaches is time
consuming. Furthermore, system-level reliability prediction in
a converter considering the wear-out failure of Semiconductor
Devices (SDs) and Capacitors (Caps) is explored in [18].
Besides the aforementioned shortcomings of MCS used in this
paper, the model uncertainties of the capacitor lifetime model
Fig. 1. Typical bathtub curve describing failure rate of an item.
are not appropriately conducted.
Also, the lifetime models provided in [17]–[21] are more
applicable for wear-out modeling. The aging failure
probability can be used for design for reliability and end-of-
life prediction in power converters. However, the system-level
design, planning, operation and maintenance of power
electronic systems require the converter availability modeling.
The converter availability depends not only on the wear-out
failure rate, but also on the failure rate of useful lifetime.
Therefore, a complete failure rate prediction within useful life
and wear-out periods is required for converter design and
operation.
In order to address the aforementioned shortcomings of the
state-of-the-art methods, this paper proposes a comprehensive
reliability prediction approach for power converters. The
proposed approach predicts the failure characteristic of a
converter within its useful and aging period according to an
applied mission profile. In the proposed approach, the constant
failure rate prediction based on the handbook estimation
method is merged with the wear-out failure estimation
Fig. 2. Failures classification based on failure causes (sources) according to
approach. Furthermore, a Stress-Strength Analysis Method
[33], [34].
(SSAM) is proposed for the wear-out failure rate prediction
employing a structural reliability concept [26]. Unlike MCS, debugging processes and they have been solved before
the proposed SSAM can quickly predict the aging probability, operating the item/system. Therefore, the item will experience
which facilitates the reliability modeling, design for reliability random chance failures within its useful lifetime. Moreover,
and reliability evaluation in large-scale power electronic based due to the aging of materials, the item may enter wear-out
power systems. phase depending on the materials strength and applied stresses
The proposed approach can facilitate reliable design and within long-term operation.
manufacturing of converters by identify its weakest links from Optimal and reliable design and operation of a converter
reliability standpoint. Furthermore, it can be used for system- depend on its hazard behavior in useful life and wear-out
level decision-making within planning, operation and phase. This is due to the fact that the long-term performance
maintenance of power electronic systems in order to enhance of the converter remarkably depends on its useful lifetime and
the overall system reliability. This is due to the fact that the its availability. During the useful lifetime, failures occur by
power electronic system reliability depends on different chance which yield a constant failure rate. The useful lifetime
factors including converter topology and its application [9], terminates once the item enters the wear-out phase where the
[27], [28] control/switching schemes [2], [8], [27], [29], [30], failure rate rises. Therefore, wear-out failure probability
operating conditions [8], [29]–[31], climate conditions [15], prediction is of paramount significance since it can affect the
[16] and so on. Therefore, during planning phase, selecting overall system life cycle and operational/ maintenance costs.
suitable converter topologies for a desired application Therefore, the design for reliability concept has introduced to
considering the climate conditions can enhance the system accurately design the components of a system to achieve a
reliability. Furthermore, appropriate control/switching desired lifetime with a certain probability. Moreover, the
schemes can extend the converter lifetime and thus the overall limiting state unavailability U for an item is defined as [32]:
system reliability during operation phase. Moreover,  ,
predicting the converter end-of-life will aid proper U = (1)
+
maintenance scheduling for different converters to retain the where λ denotes the failure rate and μ is the repair/replace rate.
system availability. In practice, the average operating time (1/λ) is much greater
In the following, the concept of reliability in power
than the average down time (1/μ) that means μ ≫ λ, and U ≈
converters is discussed in Section II. Section III presents the
λ/μ. Hence, higher failure rate causes higher unavailability and
reliability modeling within useful life. Moreover, the proposed
consequently higher system risk. Therefore, the system
SSAM is explained in Section IV. The proposed approach is
unavailability within useful lifetime should be acceptable.
exemplified by predicting the reliability of a Photovoltaic
Moreover, entering wear-out phase, the failure rate will
(PV) inverter under different operating conditions in Section
remarkably be increased, hence, the system unavailability and
V. Finally, the outcomes are summarized in Section VI.
risk will be aggravated. In such a case, suitable maintenance
Reliability of Power Converters strategies should be adopted to retain the overall system risks
in an acceptable level. Therefore, unexpected operation and
A typical hazard (failure rate) function of an item/a system maintenance costs will be induced if the failure rate of system
within its life cycle is shown in Fig. 1 including infant is not appropriately predicted. Consequently, power converter
mortality, useful life and wear-out periods. Usually the infant failures can affect the overall power system performance as
mortality failures are related to the manufacturing and well as the investment and operational costs. Hence, the failure

0885-8993 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 27,2020 at 07:25:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2020.2981933, IEEE
Transactions on Power Electronics

rate prediction within useful life and wear-out phase is random chance failure or wear-out failure is triggered.
necessary for design and operation of converters. Therefore, the total converter failure rate, λC is equal to:
General failure causes of an item can be classified into C = C −useful + C − wear (2)
random and systematic failures as shown in Fig. 2 [33], [34]. where, λC-useful is the useful life failure rate and λC-wear is the
The random failures occur at a random time resulting from one wear-out failure rate. The useful life and wear-out failure rates
or more degradation mechanisms in the hardware. These are obtained by adding the failure rate of individual
failures may be caused by human error or associated with the components, i.e., Caps and SDs as:
hardware. The hardware (physical) failures are classified into
C −useful =  Caps −useful + SD −useful (3)
random chance failure and wear-out (aging) failure [22]. The
random chance failures, also called catastrophic failures [35], C − wear =  Caps − wear + SD −wear (4)
are caused by sudden overstress, such as overcurrent In (3) and (4), it is assumed that the converter will fail if one
/overvoltage. These failures are modeled by the Exponential of the components fails, hence a series reliability network is
distribution function. Moreover, the aging failures, so-called employed to model its reliability. In the case of stand-by
gradual failures, are related to the wear-out phase of an item, systems and redundant configurations, suitable reliability
which can be modeled, e.g., by a Weibull distribution. On the modeling techniques such as Markov Process can be adopted
other hand, the systematic failures are associated in a [32]. The converter reliability is obtained by using:
deterministic way with a certain cause, which can solely be  t 
removed by a modification of the design and manufacturing R ( t ) = exp  −  C ( ) d  , (5)
processes, operational procedures or other relevant factors  0 
[33]. The systematic failures have non-physical causes, and where R(t) is the converter reliability at instant t. In the
will not re-appear if the causes are suitably corrected. following, the prediction of random chance and wear-out
Different root causes of random and systematic failures are failure rates of converters are presented.
summarized in Fig. 2 and more definition can be found in [33], Constant Failure Rate Prediction
[34]. In this paper, it is assumed that the converter is designed
perfectly, that systematic failures will never appear and the The failure rate during useful lifetime can be predicted
expert staffs are employed for operation and maintenance in considering the historical failure data within last operation of
order to avoid the random human failures. Therefore, the only the converter. The more accurate data come from the long-
likely failures, which cannot be eliminated, include the term operation under identical operating conditions. These
random chance and wear-out failures. type of data, so-called user-provided data [24], may be
In power electronic converters, following field data and obtained based on maintenance database and shutdown
industrial experiences, Capacitors (Caps) and Semiconductor reports. Moreover, in the case the reliability data are not
Devices (SD) are the two most fragile components [5], [6], available, some generic data provided in handbooks can be
[36], [37]. They are exposed to random hardware failures employed [24]. Another data source for reliability estimation
which can be single-event catastrophic failures occurred is the data prepared by the manufacturers [24]. Moreover, in
within useful lifetime and long-term wear-out failures [22], most cases, especially during the design phase of new
[23], [35], [38]–[42]. The wear-out failures, namely intrinsic technologies, these data are not available, hence, expert
failures, are originated by internal degradation of component judgment elicitation [24] could be the only option in which the
materials. Hence, they can be predicted by comparing the data of similar cases may be employed by reasonable
material mechanical strength with the applied stresses or justifications. This approach is a difficult process.
accelerated life testing. As already mentioned, the MIL-HDBK-217, Telcordia SR-
The random chance failures, which are usually extrinsic 322, Siemens SN29500, RDF-2000, IEC-TR-62380, IEC-
and caused by suddenly overstressing the components, are 61709 and FIDES [23]–[25] prepared methods and base
estimated based on field retuned data. The failure data can be failure data for components where the failure rates can be
collected and categorized following failure sources and modified according to the operating conditions. It is also
mechanisms within a long-term operation. Thereafter, a possible to use the manufacturer or user-provided data as the
complete reliability model for a specific operating condition base failure rate in order to predict the failure rate under
can be derived. This procedure requires long-term operation desired operating conditions. Moreover, the IEC-TR-62380,
data and proper classification of the failure causes and IEC-61709 and FIDES provides a general mission profile-
mechanisms under operating conditions. Using these data for based approach for electronic components operating at
the same item operating in another condition requires different conditions. According to [23]–[25], the failure rate of
reasonable justifications [24] due to the impact of operating a component can be obtained as a weighted average of failure
condition. rate in different operating phases. The failure rate of each
Meanwhile, during design and planning process, long-term phase can be predicted based on the reference failure rate
field data do not exist. Hence, the failure data of similar cases provided by manufacturer/field data/handbooks, which are
can be employed by fair justifications. In practice, the field modified according to the operating condition considering
data of similar cases can be used for obtaining a reference AFs. Moreover, the FIDES approach provides a detailed
(base) failure rate for a component under specific conditions. method for estimating the constant failure rate of electronic
Notably, this can also be provided by the manufacturer. components due to the fact that it considers the statistics of
Moreover, test data can be used for modeling the impact of possible failure causes according to the physics of failure
operating conditions on the failure rate by defining analysis [25].
Acceleration Factors (AFs) in order to model the impact of Following the FIDES approach, the failure rate of an item
operating conditions such as temperature, voltage and (λ) is predicted by using (6) [25].
humidity.  = Π PM Π Prosess Phy , (6)
This paper aims at predicting the reliability of power where,
converters considering random chance and wear-out failures Phase
 tannual 
according to the accessible failure data and models for the Phy =   8760 
i =1
Πi i , (7)
converter components. Generally, a component/system failure i

occurs once one of the failure mechanisms due to either

0885-8993 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 27,2020 at 07:25:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2020.2981933, IEEE
Transactions on Power Electronics

0.511ln( Cs )
Πi = ( Π Placement Π App Π Rugg ) , and (8)
i =  0 k Πk , (9)
k

in which, ΠPM is the impact of quality and technical control

over manufacturing, and ΠProcess models the effect of all
processes, from specification to field operation and
maintenance. The physical contribution is modeled by λPhy,
which is given in (7) considering the mission profile., where, Fig. 3. Failure probability estimation concept based on mismatch of stress
tannual is the duration of ith phase within one year. The term Πi and strength.
in (8), is the induced electrical, mechanical and thermal have cognitive and noncognitive sources including physical
overstresses. The parameters in (8) is defined in [25]. The term randomness, parameters and variable uncertainties, model
λi in (9) is the corresponding failure rate in each phase of the uncertainties, definition of quality and performance of failure,
mission profile, in which, λ0k is the base failure rate of the item, deterioration, and so on [26]. These uncertainty sources can
which can be found in the handbooks or provided by the generally be classified into two categories including aleatory
manufacturer. The AFs of Πk reflects the physical constraints and epistemic uncertainties [26], [44]. The aleatory sources are
the item experiences within operation or dormant phases. The
inherently random and non-deterministic in nature. This type
failure rate of λi is divided into thermal, case and solder joints
related, as well as humidity, and mechanical stresses. In of uncertainty is related to the physical world and cannot be
particular, the failure rate in (7) for SDs, λPhy-SD is defined as: reduced by obtaining more information and knowledge.
 0TH  Thermal  Moreover, the epistemic uncertainties are due to the
  incomplete knowledge, which can be reduced by enhancing
 + 0TCyCase  TCyCase 
Phase
t  the knowledge base [26], [44].
Phy − SD =   annual   +0TCySolderjoionts  TCySolderjoionts  (  Induced )i . (10)
i =1  8760  i
  In SDs, the wear-out failure mechanisms include bond-
 + 0 RH  RH  wire lift-off/cracking, chip solder joints cracking and
 
+  
 0 Mech Mech i baseplate solder joints cracking [22]. A device will fail due to
The failure rate of Caps is also obtained by using (11). the deformation of its structure caused by one or more failure
  Thermo −electrical 
mechanisms. The source of uncertainties in this case can be
Phase
 tannual    the strength model of device materials, characteristics of the
Phy −Cap = 0Cap     + TCy  (  Induced )i (11)
materials and applied load on the device. There are different
i =1  8760  i  
 + Mechanical i models provided in the literature for predicting the strength of
The base failure rates, λ0X and the corresponding AFs, ΠX for bond-wires and solder joints [45]. In this paper, an empirical
a failure factor X has been provided in page 120 for SDs and lifetime model is employed for predicting the reliability of the
page 138 for Caps in [25]. However, these values can be SDs as [46]:
provided by the manufacturer or predicted based on     ton 
N f = A  T j  exp   , (14)
operator/user experiences. Following the accuracy of the data,  T   1.5 
 jm 
one of the handbooks methods [23]–[25] can be employed.
where, Nf is the number of cycles to failure, ΔTj and Tj are the
The total converter failure rate during its useful lifetime
junction temperature swing and mean values, and ton is the rise
can be modeled considering the series reliability block
time of temperature cycle. A, α, β and γ are lifetime model
diagram as any individual component failure cause converter
constants, which can be obtained from aging tests [46]. In this
failure. Therefore, the converter constant failure rate, λC-useful is
model, A, α, β and γ are the epistemic sources of uncertainty
the sum of failure rate of individual components of Caps, λCaps-
in the lifetime model where their accuracy can be enhanced by
useful and SDs, λSD-useful as (3).
repeating lifetime tests. Moreover, ΔTj and Tj depend on the
This paper considers the impact of two fragile components. component electro-thermal characteristics which vary from
Notably, for more detailed analysis, the failure rate of other sample to sample due to the manufacturing uncertainties.
components provided in [25] can also be included. Moreover, Therefore, these variables cause aleatory uncertainties in the
the software reliability can also be predicted according to lifetime prediction, which cannot be reduced by collecting
IEEE Std 1633 [43]. more data.
Furthermore, the wear-out failure mechanisms of an
Wear-out Failure Rate Prediction electrolytic Caps include electrolyte vaporization and electro-
The components wear-out failure distribution can be chemical reaction [22]. Its lifetime can be modeled as [47]:
Tn −To − n2
estimated by Stress-Strength Analysis (SSA). In this approach,  Vo 
which is adopted from structural reliability [26], the Lo = Ln  2 n1
  , (15)
component strength (Resistance, R) is compared to the applied  Vn 
load (Stress, S), and hence, the performance function Z is in which, Ln denotes the nominal lifetime under nominal
expressed as: voltage Vn and nominal temperature Tn, and Lo is the capacitor
Z =R−S . (12) lifetime under operating voltage Vo and temperature To. The
Therefore, failure probability Pf is obtained by using (13). constants n1 and n2 are provided in [47]. In this model, n1, n2
Pf = Pr ( Z  0 ) , (13)
and Ln are the epistemic uncertainty sources, whereas To and
Vo are the aleatory uncertainty sources.
where Pr(·) denotes the probability of (·). As shown in Fig. 3, In the provided lifetime model for SDs and Caps, the
both stress and strength may have uncertainties in practice. epistemic uncertainties such as model constants must be
Uncertainty can be defined as knowledge incompleteness due determined by lifetime tests. The accurate reliability model
to the inherent deficiencies with acquired knowledge [26]. It requires more tests with an acceptable confidence level.
could be associated with the ambiguity and vagueness in Moreover, the aleatory uncertainties such as temperature come
defining paraments and variables of components. They may from manufacturing variations and applied mission profile.

0885-8993 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 27,2020 at 07:25:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2020.2981933, IEEE
Transactions on Power Electronics

These uncertainties must be accurately defined based on the

provided data by manufacturers and precisely electro-thermal
mapping of mission profile. After recognizing uncertainties,
the density function of stress and resistance can be identified
and consequently the component reliability can be obtained by
using (13).
The failure probability described by (13) can be obtained
by MCS [17], [18]. In practice, for large-scale power Fig. 4. Failure probability estimation concept based on proposed SSAM.
electronic based power systems, MCS is not feasible due to
the calculation burden. Hence, a First Order Reliability According to (22) and (23), the converted random variables
Method (FORM) [26] is adopted in order to find the will yield an equivalent strength and damage with the
components failure probability. stationary values which is identical to the stationary one.
The components resistance R in (14) and (15) can generally Finally, the performance function in (18) is re-defined as:
be represented as: Z ( t ) = Req − tST = g ( x1,eq ,...,xn,eq ) − tST (24)
R = g( x1 ,...,xn ) , (16)
The failure probability can be calculated by integrating the
where x1, …, xn are random variables describing the
stress function at its left-hand tail limited by the stress level as
component strength by a function of g(·). Time variant
shown in Fig. 4. This integration can be predicted by the
performance function Z can be expressed as:
FORM [26]. In this method, the mean μ and variance σ of
Z ( t ) = R − tS;t = 1,2,... , (17)
random variable Z are estimated by the first order Taylor
where, t is the multiple of period of applied stress S. In approximation as:
practice, S is not a stationary stress over the mission profile,
and it comprises of different levels of {S1, …, Sh} according to
(
Z  g  x ,...,  x
1,eq n ,eq
) − tS
T , and (25)

 g ( x1,eq ,...,xn,eq )  2
2
the applied mission profile [2]. Therefore, the performance n
function can be modified as:   
2
 x , (26)
i =1  xi ,eq  i ,eq
Z
h  
Z ( t ) =  ( Ri − tSi );t = 1,2,... , (18) where ∂ denotes a partial differential operator, and μθ and σθ
i =1
are the mean and variance of random variable θ. As already
where h is the total number of stress levels within a period of
mentioned, Z may follow the normal distribution in practice.
time, e.g., one year, and Ri = g(x1,i, …, xn,i) is the component
Therefore, the failure probability is obtained as:
strength due to the applied stress of Si. The failure probability
  
at time t can be calculated by substituting (18) in (13). In Pf ( t )  Φ  − Z  (27)
practice, h is a quite large number if an annual mission profile  Z 
is considered. Thereby, the random variables Z may follow the where Φ(·) is the standard normal distribution function. The
Normal Distribution following the central limit theorem [26]. component reliability R(t) = 1 – Pf (t). Finally, the wear-out
Solving (13) requires multiple integration to obtain the failure rate of component X is calculated by:
1 dR ( t )
failure probability, which can be a difficult task. To avoid X − wear ( t ) = − , (28)
computational difficulties, the non-stationary stress should be R ( t ) dt
transformed into a stationary one. First, the impact of different where, d is a differential operator.
stress levels of the component can be defined as a Damage (D),
which is defined as: Case Study on a PV Inverter
h
Si
D= . (19) The reliability prediction procedure in power converters
i =1 Ri has been discussed in previous conventional sections. The It is
Next, the stationary stress and equivalent resistance should be highlighted that the converter reliability depends on both
defined in such a way that the resultant damage to be identical random chance failure and wear-out failure. Furthermore, both
to the non-stationary one obtained by (19) [48]. The total stress failure types can be affected by the operating and climate
conditions, converter topology, control/switching scheme and
ST over a mission profile is:
h
so on. In this section, the proposed method is exemplified for
ST =  Si , (20) a PV inverter in order to show the impact of operating
i =1 conditions (which is associated to the solar irradiance) and
and, the resultant strength RT due to the applied stresses is: climate conditions (here the ambient temperature). The
ST inverter reliability is predicted considering both random
RT = , (21) chance and wear-out failures under two mission profiles.
D
Furthermore, the effectiveness of the proposed SSAM for
Furthermore, the equivalent random variables {x1,eq, …, xn,eq}
predicting the aging failure rate is evaluated and compered to
can be determined to obtain the same damage as for the set of the conventual Monte Cardo simulation-based approach.
random variables {x1,i, …, xn,i} i = 1, …, h. Therefore, the Moreover, it identifies the weakest links of the converter
equivalent of the ith random variable, except the kth one, is operating under different mission profiles, that is useful to
defined as its average value using (22). enhance the converter reliability during design process. The
1 h inverter reliability function is beneficial for proper decision-
i =1:h xi ,e q
ik
 xi ,h
h j =1
(22)
making in planning of the power systems for cost analysis and
maintenance scheduling based on the predicted failure rate and
The equivalent of the kth random variable is obtained by (23).
end-of-life of inverter.
xk ,e q g −1 ( RT ) x(h) ,h k (23) This case study shows the detailed analysis of the proposed
i ,eq

reliability prediction approach in a PV converter, that can be

applied for different converters with different applications.
The structure of the double-stage PV inverter is shown in Fig.
0885-8993 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 27,2020 at 07:25:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2020.2981933, IEEE
Transactions on Power Electronics

Fig. 5. Structure of the single-phase double-stage PV inverter.

TABLE I. INVERTER COMPONENTS PARAMETERS AND LIFETIME MODEL.
Parameter Value Parameter Value
L 2 mH A 9.34E14 ± 5%
Cb 120 μF α -4.416 ± 5%
Ci 3×390 μF β 1285 ± 5%
Qb IGB10N60T γ 0.3 ± 5%
Qi GB15N60T n1 10 ± 5%
Db IDV20E65D1 n2 3 ± 5%
Di IDV20E65D1 Tn 105oC Fig. 7. Inverter switch wear-out (a) pdf, (b) cdf, using Monte Carlo analysis
fsw 20 kHz Vn 450 V and the proposed SSAM under the mission profile of Location B.
Vdc 400 V Vac 230 V, 50 Hz

Fig. 8. Inverter capacitor wear-out (a) pdf, (b) cdf, using Monte Carlo
analysis and the proposed SSAM under the mission profile of Location B.

computer with Intel (R) Core (TM) i7-7600U CPU @ 2.8

GHz and 8 GB memory.
Fig. 6. Annual solar irradiance and ambient temperature for (a) Location A, The performance of the proposed approach is further
(b) Location B. evaluated by predicting the inverter capacitor and the results
5. The inverter includes a boost stage as a maximum power are compared to the MCS based approach. As shown in Fig.
point tracker and a single-phase inverter for connecting a 2.5 8(a), the failure pdf of the proposed approach is asymptotically
kW PV array to the grid. The measured mission profiles of the following the MCS results. Moreover, the predicted B10
solar irradiance and ambient temperature of two different lifetime has 1/80 = 1.2% error as shown in Fig. 8(b), while the
locations are used for the analysis as shown in Fig. 6. The computational effort is reduced by 1.50/0.02 = 75. The
converter parameters and specifications are provided in induced error is due to the fact that the proposed approach
TABLE I. relies on a first order estimation of strength function in (25)
This section includes three sub-sections. The first sub- and (26). For accurate results, the higher order approximations
section presents the effectiveness of the proposed SSAM for can be employed.
wear-out failure probability prediction. Moreover, the inverter The presented cases evaluate the performance of the
reliability is estimated in the second sub-section employing proposed SSAM compared to the conventional MCS-based
the proposed reliability prediction method. Finally, the last method from accuracy and performance standpoints. Follow-
sub-section demonstrates the operating conditions impacts on ing the obtained results, the proposed SSAM can remarkably
the inverter reliability by the help of experimental tests. reduce the calculation efforts for large-scale power electronic
based power systems with a huge number of converters.
A. Effectiveness of the proposed SSAM Moreover, in design for reliability applications, running MCS
The performance of the proposed SSAM is examined by for each iteration of design procedure is time consuming,
predicting the wear-out failure probability of the inverter while the proposed approach can facilitate this process as well.
switch and capacitor operating under mission profile of B. Comprehensive reliability prediction
Location B shown in Fig. 6. The lifetime model parameters
and corresponding uncertainties are summarized in TABLE I. In this sub-section, the inverter reliability is predicted
Moreover, the MCS is run for 10,000 samples of uncertain under both mission profiles shown in Fig. 6. First, the failure
parameters. The switch wear-out failure probability density rate of converter components within their useful lifetime is
function (pdf) and cumulative distribution function (cdf) estimated based on FIDES approach and the results are shown
estimated by both approaches are shown in Fig. 7. As it can be in Fig. 9. The components failure rate under mission profile B
seen in Fig. 7(a), the predicted failure pdf both approaches are is much higher than mission profile of Location A as shown in
almost the same. The predicted B10 lifetime has a negligible Fig. 9. Thereby, the components constant failure rate
error (0.12/22 = 0.6%) as shown in Fig. 7(b), while the significantly depends on the operating conditions. Moreover,
simulation burden is reduced by 1.85/0.02 = 60 times by according to Fig. 9, the boost capacitor (Cb) and diode (Db)
employing the proposed approach. Notably, the simulations with inverter capacitor (Ci) are the weakest components under
have been run in MATLAB environment on a personal mission profile A and the corresponding failure rates are
0885-8993 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 27,2020 at 07:25:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2020.2981933, IEEE
Transactions on Power Electronics

almost identical. However, the failure rate of the boost diode

(Db) is the most stressed component followed by boost switch
(Qb) and inverter switch (Qi) under mission profile B as shown
in Fig. 9. Therefore, not only the failure rate of components,
but also the weakest links of converter significantly depend on
the operating conditions. This is due to the fact that the
induced thermal cycles and the average temperature are not
identical for both mission profiles. Therefore, following (10)
and (11), the resultant failure rate on different components will
be different.
In the next step, the wear-out failure rate of converter
components is estimated and shown in Fig. 10(a) and (b)
respectively for mission profiles of Locations A and B.
Following the obtained results shown in Fig. 10(a), the
inverter switch (Qi) and capacitor (Ci) are the most fragile
components under mission profile A from wear-out point of Fig. 9. Constant failure rate of individual components of converter
under the two mission profiles according to FIDES.
view. However, the inverter switch (Qi) is the only fragile
component under mission profile B as shown in Fig. 10(b).
Moreover, the components operating in Location B is more
prone to wear-out failures compared to the operating condition
in Location A as shown in Fig. 10. These results show that the
wear-out failure also depends on the mission profiles.
Furthermore, the inverter switch (Qi) is the vulnerable
component in both cases, while the inverter capacitor (C i) is
also a weak link in Location A. As a result, inverter switch and
capacitor limit the converter useful life expectancy under
mission profile A, and in the case of mission profile B, the
inverter switch is the only player.
The analysis shows that the stress on the components
within useful lifetime is different from that of within wear-out
phase as shown in Fig. 9 and Fig. 10. This outcome could be
beneficial for appropriate maintenance planning during useful
lifetime depending on the operating conditions. Moreover, the Fig. 10. Wear-out failure rate of individual components of converter
converter availability can be improved by re-designing the under the mission profile of (a) Location A and (b) Location B.
weakest components during useful life. Furthermore, the
converter can be properly designed to achieve a desired The converter reliability due to components aging under
lifetime by predicting its components end-of-life and wear-out both mission profiles is shown in Fig. 11(b) and Fig. 12(b)
behavior. The wear-out analysis is also beneficial for system with blue graph. The B1 lifetime of the converter due to the
level preventive maintenance, where the converter wear-out failures in Location A is 60 years, while it is 9 years
unavailability rises over an undesired value. for location B. This fact is because of the different stress levels
In the following, the total constant failure rate of the induced by the mission profiles. The total reliability of the
converter is calculated by summing the failure rate of all converter is shown in Fig. 11(b) and Fig. 12(b) with a red
components. The total constant failure rate is shown with graph. The total B1 lifetime of converter under mission profile
green-dashed line in Fig. 11(a) for Location A and in Fig. A is 18 years and under mission profile B, it is 4 years. These
12(a) for Location B. The total wear-out failure rate under both results show that the converter design based on the wear-out
mission profiles are also shown in Fig. 11(a) and Fig. 12(a) failure may introduce an undesired consequence, hence the
with a blue line. The total failure rate of the converter is shown constant failure rate should also be considered in the design
by red line in Fig. 11(a) and Fig. 12(a), which is in fact the procedure. Moreover, the obtained results show that unlike
sum of constant and wear-out failure rates. IET TR-62380, the life expectancy remarkably depends on the
mission profiles.

Fig. 11. Failure rate and reliability of PV inverter under mission profile of Fig. 12. Failure rate and reliability of PV inverter under mission profile of
location A. location B.

0885-8993 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 27,2020 at 07:25:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2020.2981933, IEEE
Transactions on Power Electronics

Fig. 13. Photograph of the implemented dc-dc boost converter.

C. Operation impact on device temperature

The temperature of the device is the main factor limiting
its lifetime. It depends on the operating condition and mission
profiles and this fact was the motivation to include the mission
profile analysis in the reliability studies. Therefore, the impact
of operating conditions on the temperature of different
components in the converter is demonstrated in this sub-
section. A photograph of the test prototype is shown in Fig.
13, where the junction temperature of power module and the
hotspot temperature of the capacitor are measured under Fig. 14. Obtained experimental results. (a) The converter components
temperature at different loading condition. Thermal image of (b) Qb & Db
different loading conditions. temperature at 1.8 kW, and (c) Qi & Di temperature at 1.8 kW.
The converter is tested under 0.8 kW, 1.8 kW, and 3.2 kW,
load power and the components temperatures are measured as Conclusion
shown in Fig. 14(a). As shown in Fig. 14(a), the components
temperature depends on the operating conditions. Moreover, This paper has proposed a guideline for reliability
the temperature variation in terms of converter loading is not prediction in power electronic converters. The failure
linear. As shown in Fig. 14(a), the temperature rises at high characteristics within useful life is estimated based on generic
handbook-provided data. Moreover, the wear-out failure rate
loads, e.g., 2-3.2 kW is greater than lower loads, i.e., 0.8-1.8
is predicted based on the proposed Stress-Strength Analysis
kW. This shows that, for example, the Qi failure rate operating Method (SSAM) employing the concept of structural
at low powers will be less than operating at high load powers. reliability, where the physics of failures are taken into account.
This is shown in Fig. 10, where the failure rate of Qi in location The proposed reliability model can be used for optimal design
B is higher than location A. This is due to the different load and manufacturing of the converters as well as for system-
variations in location A and B as shown Fig. 6, where within level planning, operation and maintenance of power electronic
months from 8 to 12 in location A, the loading is low and systems.
hence the failure rate will be low. Moreover, comparing the The proposed method is exemplified for a photovoltaic
temperature difference of Db with Di (and Qb with Di) shows inverter under two climate conditions. The obtained results
that at low powers the Di has the greater temperature, while at show that the proposed SSAM for aging failure prediction
high power Db (and Qb) has the greater temperature. This introduces 60⁓70 times lower calculation burden compared to
shows that different components may limit the converter the conventional Monte Carlo Simulation based approaches.
This can facilitate the power electronic-based power systems
reliability depending on the operating condition. Furthermore,
reliability modeling and evaluation with a large number of
according to Fig. 14(a), the temperature level of Qi is higher
aging-prone components. Furthermore, the analysis shows
than other components, which may make it dominant that the individual components failure rate and the weakest
component limiting the converter lifetime. links within useful life and aging period remarkably depend
In general, the temperature rise, temperature difference and on the applied mission profile. Moreover, the converter
temperature level will affect the components reliability and weakest link in useful life may be different from the wear-out
consequently limiting the converter lifetime. As a result, the phase according to the employed reliability models provided
reliability prediction approaches, such as MIL-HDBK-217, in the literature. Therefore, strengthening the converter in both
which rely on the rated power, cannot accurately estimate the phases requires accurate reliability modeling according to the
converter lifetime. Moreover, it cannot be used to identify the applied mission profile. Finally, the design for reliability
weakest links of the converter, and hence, improve its based on the wear-out failure probability introduces higher
reliability during design and manufacturing. Hence, the difference with the case the total failure probability is
mission profile analysis is required for reliability modeling in considered. For instance, the B1 lifetime of inverter under
mission profile A changed form 60 years to 18 years when
converters. The thermal stress distribution over the
both failures are modeled. This fact may cause an unreliable
semiconductor devices for both boost and inverter is further
design of a converter, hence, a complete reliability model
illustrated in Fig. 14(b & c) operating under 1.8 kW. According within useful and wear-out phases is required for reliable
to these results, the inverter switch, Qi has the highest design of a converter.
temperature. This obtained result is identical with the SSAM
results provided in Fig. 10. References
[1] J. Falck, C. Felgemacher, A. Rojko, M. Liserre, and P. Zacharias,
“Reliability of Power Electronic Systems: An Industry Perspective,”
IEEE Ind. Electron. Mag., vol. 12, no. 2, pp. 24–35, Jun. 2018.
[2] S. Peyghami, H. Wang, P. Davari, and F. Blaabjerg, “Mission Profile

0885-8993 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 27,2020 at 07:25:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2020.2981933, IEEE
Transactions on Power Electronics

Based System-Level Reliability Analysis in DC Microgrids,” IEEE Reliability Enhancement of DC/DC Stage in a Single-Phase PV
Trans. Ind. Appl., vol. 55, no. 5, pp. 5055–5067, 2019. Inverter,” Microelectron. Reliab., vol. 88–90, no. September, pp. 1030–
[3] F. Blaabjerg, Y. Yang, D. Yang, and X. Wang, “Distributed Power- 1035, 2018.
Generation Systems and Protection,” Proceedings of the IEEE, vol. 105, [28] S. Peyghami, A. Abdelhakim, P. Davari, and F. Blaabjerg, “Reliability
no. 7, pp. 1311–1331, Jul. 2017. Assessment of Single-Phase PV Inverters,” in Proc. IEEE ECCE ASIA
[4] H. E. Johan, G. W. Steven, and H. Ramtin, “Third EGrid Workshop (ICPE), 2019, pp. 1–7.
Maps the Grid of the Future: Attendees Engage to Examine the Role of [29] V. Raveendran, M. Andresen, and M. Liserre, “Reliability Oriented
Power Electronic Applications in Modern Electric Power Systems,” Control of DC/DC Converters for More Electric Aircraft,” in Proc.
IEEE Power Electron. Mag., vol. 6, no. 1, pp. 48–55, 2019. IEEE ISIE, 2018, pp. 1352–1358.
[5] S. Yang, A. Bryant, P. Mawby, D. Xiang, L. Ran, and P. Tavner, “An [30] K. Ma, M. Liserre, and F. Blaabjerg, “Reactive Power Influence on the
Industry-Based Survey of Reliability in Power Electronic Converters,” Thermal Cycling of Multi-MW Wind Power Inverter,” IEEE Trans. Ind.
IEEE Trans. Ind. Appl., vol. 47, no. 3, pp. 1441–1451, May 2011. Appl., vol. 49, no. 2, pp. 922–930, Mar. 2013.
[6] Y. Song and B. Wang, “Survey on Reliability of Power Electronic [31] S. Peyghami, P. Davari, D. Zhou, M. F-Firuzabad, and F. Blaabjerg,
Systems,” IEEE Trans. Power Electron., vol. 28, no. 1, pp. 591–604, “Wear-Out Failure of a Power Electronic Converter Under Inversion and
Jan. 2013. Rectification Modes,” in Proc. IEEE ECCE, 2019, pp. 1598–1604.
[7] S. Peyghami, F. Blaabjerg, and P. Palensky, “Incorporating Power [32] R. Billinton and R. Allan, “Reliability Evaluation of Engineering
Electronic Converters Reliability into Modern Power System Reliability Systems.” New York: Plenum press, 1992.
Analysis,” IEEE J. Emerg. Sel. Top. Power Electron., no. DOI [33] “IEC 61508: Functional Safety of Electrical/Electronic/Programmable
10.1109/JESTPE.2020.2967216, 2020. Electronic Safety-Related Systems - Part 4: Definitions and
[8] S. Peyghami, P. Davari, and F. Blaabjerg, “System-Level Reliability- Abbreviations,” 2010.
Oriented Power Sharing Strategy for DC Power Systems,” IEEE Trans. [34] M. Rausand, “Reliability of Safety-Critical Systems: Theory and
Ind. Appl., vol. 55, no. 5, pp. 4865–4875, 2019. Applications.” John Wiley & Sons, 2014.
[9] S. Peyghami, P. Davari, H. Wang, and F. Blaabjerg, “The Impact of [35] R. Wu, F. Blaabjerg, H. Wang, M. Liserre, and F. Iannuzzo,
Topology and Mission Profile on the Reliability of Boost-Type “Catastrophic Failure and Fault-Tolerant Design of IGBT Power
Converters in PV Applications,” in Proc. IEEE COMPEL, 2018, pp. 1– Electronic Converters - An Overview,” in Proc. IEEE IECON, 2013, pp.
8. 507–513.
[10] P. Tu, S. Member, S. Yang, and S. Member, “Reliability and Cost Based [36] F. Costa, “Failure Analysis of the Dc-Dc Converter: A Comprehensive
Redundancy Design for Modular Multilevel Converter,” IEEE Trans. Survey of Faults and Solutions for Improving Reliability,” IEEE Power
Ind. Electron., vol. 66, no. 3, pp. 2333–2342, 2019. Electron. Mag., vol. 5, no. December, pp. 42–51, 2018.
[11] F. H. Aghdam, S. Member, and M. Abapour, “Reliability and Cost [37] H. Wang, K. Ma, and F. Blaabjerg, “Design for Reliability of Power
Analysis of Multistage Boost Converters Connected to PV Panels,” Electronic Systems,” in Proc. IEEE IECON, 2012, pp. 33–44.
IEEE J. Photovoltaics, vol. 6, no. 4, pp. 981–989, Jul. 2020. [38] H. Wang and F. Blaabjerg, “Reliability of Capacitors for DC-Link
[12] S. Xu, H. Chen, S. Member, F. Dong, J. Yang, and S. L. D. Old, Applications in Power Electronic Converters - An Overview,” IEEE
“Reliability Analysis on Power Converter of Switched Reluctance Trans. Ind. Appl., vol. 50, no. 5, pp. 3569–3578, Sep. 2014.
Machine System under Different Control Strategies,” IEEE Trans. Ind. [39] M. Pecht and J. Gu, “Physics-of-Failure-Based Prognostics for
Electron., vol. 66, no. 8, pp. 6570–6580, 2019. Electronic Products,” Trans. Inst. Meas. Control, vol. 31, no. 3, pp. 309–
[13] S. Peyghami, P. Davari, H. Wang, and F. Blaabjerg, “System-Level 322, 2009.
Reliability Enhancement of DC/DC Stage in a Single-Phase PV [40] N. Degrenne, J. Ewanchuk, E. David, R. Boldyrjew, S. Mollov, M. E.
Inverter,” Microelectron. Reliab., vol. 88–90, pp. 1030–1035, Sep. R, and D. C. Europe, “A Review of Prognostics and Health Management
2018. for Power Semiconductor Modules,” in Proc. Annual Conference of the
[14] V. Raveendran, M. Andresen, and M. Liserre, “Improving Onboard Prognostics and Health Management Society, 2015, pp. 1–11.
Converter Reliability For More Electric Aircraft With Lifetime-Based [41] Y. Luo, F. Xiao, B. Wang, and B. Liu, “Failure Analysis of Power
Control,” IEEE Trans. Ind. Electron., pp. 1–10, 2019. Electronic Devices and Their Applications under Extreme Conditions,”
[15] S. E. De Le, H. Calleja, S. Member, F. Chan, H. R. Jim, S. E. De Léon- Chinese J. Electr. Eng., vol. 2, no. 1, pp. 91–100, 2016.
Aldaco, H. Calleja, F. Chan, and H. R. Jiḿenez-Grajales, “Effect of the [42] H. Oh, B. Han, P. McCluskey, C. Han, and B. D. Youn, “Physics-of-
Mission Profile on the Reliability of a Power Converter Aimed at Failure, Condition Monitoring, and Prognostics of Insulated Gate
Photovoltaic Applications-A Case Study,” IEEE Trans. Power Bipolar Transistor Modules: A Review,” IEEE Trans. Power Electron.,
Electron., vol. 28, no. 6, pp. 2998–3007, 2013. vol. 30, no. 5, pp. 2413–2426, May 2015.
[16] S. E. De Le and H. Calleja, “Reliability and Mission Profiles of [43] IEEE Reliability Society, “IEEE Recommended Practice on Software
Photovoltaic Systems : A FIDES Approach,” IEEE Trans. Power Reliability,” 2016.
Electron., vol. 30, no. 5, pp. 2578–2586, 2015. [44] A. Der Kiureghian and O. Ditlevsen, “Aleatory or Epistemic? Does It
[17] P. D. Reigosa, H. Wang, Y. Yang, F. Blaabjerg, P. D. Reigosa, H. Wang, Matter?,” Struct. Saf., vol. 31, no. 2, pp. 105–112, Mar. 2009.
Y. Yang, and F. Blaabjerg, “Prediction of Bond Wire Fatigue of IGBTs [45] H. S. Chung, H. Wang, F. Blaabjerg, and M. Pecht, “Reliability of Power
in a PV Inverter under a Long-Term Operation,” IEEE Trans. Power Electronic Converter Systems,” First Edi. London: IET, 2016.
Electron., vol. 31, no. 10, pp. 3052–3059, Mar. 2016. [46] R. Bayerer, T. Herrmann, T. Licht, J. Lutz, and M. Feller, “Model for
[18] D. Zhou, H. Wang, and F. Blaabjerg, “Mission Profile Based System- Power Cycling Lifetime of IGBT Modules - Various Factors Influencing
Level Reliability Analysis of DC/DC Converters for a Backup Power Lifetime,” in Proc. IEEE CIPS, 2008, pp. 1–6.
Application,” IEEE Trans. Power Electron., vol. 33, no. 9, pp. 8030– [47] A. Albertsen, “Electrolytic Capacitor Lifetime Estimation,” JIANGHAI
8039, 2018. Eur. GmbH, pp. 1–13, 2010.
[19] K. Ma, M. Liserre, F. Blaabjerg, and T. Kerekes, “Thermal Loading and [48] J. McPherson, “Reliability Physics and Engineering,” 2nd ed.
Lifetime Estimation for Power Device Considering Mission Profiles in Switzerland: Springer Int., 2013.
Wind Power Converter,” IEEE Trans. Power Electron., vol. 30, no. 2,
pp. 590–602, Feb. 2015.
[20] F. Blaabjerg, K. Ma, D. Zhou, and Y. Yang, “Mission Profile-Oriented
Reliability Design in Wind Turbine and Photovoltaic Systems,” in
Reliability of Power Electronic Converter Systems, First Edit.,
Institution of Engineering and Technology, 2015, pp. 355–390.
[21] M. Ciappa, F. Carbognani, P. Cova, and W. Fichtner, “Lifetime
Prediction and Design of Reliability Tests for High-Power Devices in
Automotive Applications,” IEEE Int. Reliab. Phys. Symp. Proc., vol.
2003-Jan, no. 4, pp. 523–528, 2003.
[22] S. Peyghami, Z. Wang, and F. Blaabjerg, “Reliability Modeling of
Power Electronic Converters: A General Approach,” in Proc. IEEE
COMPEL, 2019, pp. 1–7.
[23] “IEC TR 62380: Reliability Data Handbook-Universal Model for
Reliability Prediction of Electronics Components, PCBs and
Equipment,” 2006.
[24] “IEC 61709 (2017): Electric Components - Reliability - Reference
Conditions for Failure Rates and Stress Models for Conversion,” 2017.
[25] “FIDES Guide 2009 Edition: A Reliability Methodology for Electronic
Systems,” 2010. [Online]. Available: www.fides-reliability .org.
[Accessed: 02-Feb-2019].
[26] B. M. Ayyub and R. H. McCuen, “Probability, Statistics, and Reliability
for Engineers and Scientists,” 3rd ed. Taylor & Francis Group, 2015.
[27] S. Peyghami, P. Davari, H. Wang, and F. Blaabjerg, “System-Level
0885-8993 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 27,2020 at 07:25:54 UTC from IEEE Xplore. Restrictions apply.