Beruflich Dokumente
Kultur Dokumente
org
Published in IET Power Electronics
Received on 17th January 2011
Revised on 20th February 2012
doi: 10.1049/iet-pel.2011.0018
ISSN 1755-4535
Abstract: This study presents a novel research methodology and approach for diagnosing pertinent failures for power electronic
converters. The concept of the generalised connection matrix for a sneak circuit analysis is used in this research. A novel
modification to this process shown in this study produces a usable feature that provides unique signatures under fault
scenarios and could be used to diagnose faults. The novelty lies in using component currents to form the generalised
connection matrix. This study presents the results obtained using data collected from fault scenarios on a hardware set-up for
the dc – dc buck converter.
814 IET Power Electron., 2012, Vol. 5, Iss. 6, pp. 813 –826
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-pel.2011.0018
www.ietdl.org
2 Related research Filter capacitors have been the focus of component-level
fault diagnosis apart from previously mentioned literature
2.1 System-level diagnostics, reconfiguration and for semiconductor switch open- and short-circuit faults.
supporting research Capacitor failures and their diagnosis are addressed in
[39 – 41]. A strategy is explained in [42] to find the
The move towards a dc distribution system places emphasis condition of an aluminium electrolytic capacitor based on
on a greater use of power electronics equipment as factors such as equivalent series resistance, capacitance and
described in Section 1. Advances in power electronics make dissipation factor values, which change with temperature,
it feasible and possible to deploy the envisioned power frequency, current ripple as well as ageing. A review of
distribution architecture. The challenges of adopting a dc intelligent techniques applied for fault diagnosis in power
power distribution system for ships are highlighted by electronics is presented in [43]. A model-based diagnostic
Baran and Mahajan [16]. Stability issues relating to the system trained with the machine learning technology for
proposed dc architecture were discussed in [17] and a inverters feeding electric drives is presented in [44].
security assessment is reported by Momoh et al. [18].
Momoh and Ishola-Salawu [19] discuss a neural network
2.3 Concept of fault accommodation
application for arcing fault detection. Butler-Purry [20],
Butler-Purry and Srivastava [21] and Butler-Purry and Fault accommodation is a related area of research which takes
Sarma [22] report the use of multi-agent systems, expert the role of automated fault diagnosis a step further [45, 46]. It
systems and self-healing aspects for the SPS. Agent-based could be viewed as a type of fault tolerance wherein the
reconfiguration is also reported by Baran and Mahajan [23]. system is reconfigured to operate at the best possible
efficiency even with the presence of one or more faults.
2.2 Power electronics development and device/ Although fault tolerance could be achieved in its simplest
component level diagnostic research form using redundancy, fault accommodation in theory is
expected to adapt controller parameters or reconfiguration
The field of high-power drives has been an active research and of the system to avoid and/or minimise the consequences of
development area of power electronic application in the last a fault where performance may degrade. A simple well-
decade. Many industrial processes have increased power- established way of fault accommodation is based on pre-
level needs, giving rise to the development of new power designed controllers, which are trained offline in known and
semiconductors, converter topologies and control methods. expected failure modes. Thereafter, appropriate online
A survey of the most important converter topologies with failure accommodation techniques are suggested by a fault
related control schemes has been presented in [24]. diagnosis system which analyses fault patterns online.
Research has also focused on associated fault diagnosis for Further related work on this methodology can be found in
power electronic equipment. A variety of techniques have [47 – 50]. Similar to fault diagnosis research activities, most
been proposed and employed for diagnosing faults in the of the fault accommodation schemes are primarily
different types of power converters and their components. developed based on the powerful and well-understood
The techniques range from comparing measured values of linear control methodology. Fig. 3 shows the various fault
input and output quantities to detect discrepancies; Artificial accommodation techniques.
Intelligence (AI)-based techniques like neural networks,
fuzzy systems; statistical techniques like Fourier transforms 2.4 Reliability analyses
and wavelets and model-based methods. Furthermore, the
diagnostics proposed include higher level device diagnosis As mentioned earlier, a lack of benchmark systems makes it
as well at the internal component level. vital to understand various pertinent faults and their
Literature exists for voltage–source inverter fault diagnostics associated risks. Following this process, research diagnosing
using normalised dc components [25], open circuit fault these faults can be undertaken. Owing to this pressing need
detection by pole-voltage measurement [26] and also using of understanding system failure at all levels, attention was
output voltage frequency analysis [27]. Dc link current shapes turned to reliability analysis techniques.
were utilised to diagnose faults in zero-voltage switching Reliability analysis techniques such as FMEA can be used
converters [28]. For multi-level inverter a fuzzy-based to support reliability, maintainability, testability, safety and
diagnosis approach is presented in [29], whereas an AI-based logistics analyses. When performed in an accurate and
diagnosis and reconfiguration technique is reported in timely fashion, FMEA information can be used to aid the
[30]. Application of wavelets to detect faults in power systems design of test systems, the development of trouble-shooting
and conversion modules has been demonstrated in [31, 32]. procedures, the planning of scheduled maintenance and
An approach using rotor current space vector angle patterns to the development of integrated diagnostic capabilities. In
detect short and open circuit faults for rectifiers in a sub- addition, an effective FMEA presents an examination of a
synchronous cascade drive is detailed in [33]. Design of system’s strengths and weaknesses [51 – 54]. The two
ground fault detection and overvoltage limiting for three- approaches to FMEA, functional FMEA and hardware
level pulse width modulation (PWM) inverter is covered in FMEA [5], were used in this research and their application
[34–36]. Further, Rothenhagen and Fuchs [37] evaluate is reported in [1].
different fault diagnosis strategies for a voltage–source Another reliability analysis method called SCA was studied
inverter-fed variable speed drive comparing MATLAB- to be utilised in this research. SCA is conventionally used
SIMULINK results with an experimental setup. Here, a during design phases of electronic circuits to help detect
comparison is made among fault-detection techniques such as unwanted current paths so that the circuit could be
Park’s vector approach, slope method and variations of using redesigned if necessary. Although designing a new
direct current methods. A comprehensive comparison and electrical circuit is not a part of this research, conventional
evaluation of over a total of 30 diagnostic methods mainly for SCA was modified to produce diagnostic features able to
IGBT open-circuit and short-circuit faults are provided in [38]. differentiate between different fault states for the dc – dc
816 IET Power Electron., 2012, Vol. 5, Iss. 6, pp. 813 –826
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-pel.2011.0018
www.ietdl.org
the design phase of a device in order to modify a design
if a sneak path is identified. This feature is not utilised in
this research. Instead, the novel modification applied
yields numeric values or signatures from the determinant
of the generalised connection matrix that are shown to
be useful for differentiating between various operating
scenarios.
been elaborated in later sections. This approach is shown in Fig. 6 shows the dc – dc buck converter circuit used for the
Fig. 5. analysis.
The PART algorithm is based on the assumption that the
model equations of PART (a large-scale singularly 5.2 Fault cases studied
perturbed system of differential equations having a reset
mechanism) have regular computational performance. The The common failure modes for components like
PART neural network developed by Cao and, rule-based semiconductor switches and diodes are typically open and
system developed by Wu have been shown to be effective short circuits. Similar failure modes exist for capacitors and
in clustering data sets in high-dimensional spaces. The inductors. Electrolytic capacitors display electrolyte leakage
proofs and simulation results are available for further and dielectric breakdown, whereas inductors may suffer
reading in [56]. The PART functionality in Weka was inter-turn winding faults. Table 1 shows fault scenarios
chosen as it produces IF – ELSE rules that can be easily conducted for the buck converter considering single and
read and interpreted as opposed to other formats such as multiple fault cases. The notion of incipient capacitor or
decision trees produced by a similar classification algorithm inductor failure was simulated on hardware by manually
called J48 (which is an extension of C4.5). reducing the net capacitance and inductance, respectively.
4.1 SCA definition The buck converter circuit has two distinct states when the
switch S1 is ON and when it is OFF. The equivalent
SCA [58 – 61] is a functional reliability analysis technique circuits during these two states are shown in Figs. 7a and
with the potential of detecting unintended and thereby 8a. The modified directed graphs are constructed for the
undesirable system operation. A sneak circuit is a latent two states separately which are shown in Figs. 7b and 8b.
path or condition that inhibits the desired function or
causes undesired operation to occur. In contrast to failure
effects, a sneak circuit does not require a component
failure to occur. SCA is a unique method of evaluating the
electrical circuit topology to detect specific patterns which
are characteristic of sneak paths. A sneak condition or path
is present in a circuit, as designed, but may not always be
active.
6 Results
6.1 SIMULINK circuit
818 IET Power Electron., 2012, Vol. 5, Iss. 6, pp. 813 –826
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-pel.2011.0018
www.ietdl.org
For both circuits, the data for fault and no-fault cases were
collected at 10 ns sampling rate over a 1-ms time period
(thus 100 000 samples). Using (3) and (4), the features
|S1off| and |S1on| are computed for the respective cases.
There are then 12 data sets, six for |S1off| (five fault
cases + no fault case) and six for |S1on|. Each set is
separately processed by the classification algorithm to yield
two results.
These data used in the classifier are first divided into ranges
and windows. The range variable reflects the total number of
Fig. 10 Hardware test set-up for buck converter circuit samples considered by the classifier at one time and is set
Table 2 Comparison of waveforms generated by the model and hardware circuit with fault detection results on hardware data
Semiconductor switch (S1) open Semiconductor switch (S1) Inductor fault case: input current Capacitor fault case: input current
circuit case: input current and short circuit case: input current and output voltage and output voltage
output voltage and output voltage
semiconductor switch (S1) open semiconductor switch (S1) short inductor fault case: output capacitor fault case: output voltage
circuit case: output voltage with circuit case: output voltage with voltage with fault indication with fault indication
fault indication fault indication
continuous fault indication fault indication approximately fault indication approximately fault indication approximately
approximately 0.3 ms after fault 0.4 ms after fault introduction 0.2 ms after fault introduction 0.17 ms after fault introduction
introduction
820 IET Power Electron., 2012, Vol. 5, Iss. 6, pp. 813 –826
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-pel.2011.0018
www.ietdl.org
Table 3 Rules derived using S1-off equation for classification of hardware data
Rules 1–5 Rules 6 –10 Rules 11– 15 Rules 16– 20 Rules 21 –23
IF Value 8 . 2500.26 AND IF Value 2 . 2349.08 AND IF Value 2 . 2341.57 IF Value 1 . 2349.08 ANDIF Value 8 ≤ 2 500.26 AND
Value 5 ≤ 2 350.48 AND Value 1 . 2341.57 AND THEN data is of type ‘d’ Value 8 ≤ 0 AND Value 4 . 2683.15 AND
Value 1 . 2698.16 AND Value 1 ≤ 2 224.59 Value 7 . 2334.53 AND Value 2 ≤ 2 806.82
Value 9 . 2358.19 AND THEN data is of type ‘c’ Value 1 ≤ 2224.59 AND THEN data is of type ‘f’
Value 8 ≤ 2 212.37 AND Value 3 ≤ 2341.57 AND
Value 4 . 2500.26 Value 4 ≤ 2218.48 AND
THEN data is of type ‘a’ Value 6 ≤ 2334.53
THEN data is of type ‘e’
IF Value 2 ≤ 2665.33 AND IF Value 3 . 2332.666 ANDIF Value 8 ≤ 2 212.37 AND IF Value 1 . 2349.08 ANDIF Value 5 ≤ 2 500.26 AND
Value 2 . 2727.5238 Value 3 ≤ 2218.4866 Value 2 ≤ 2 500.2624 Value 8 ≤ 0 AND Value 2 . 2619.93
THEN data is of type ‘a’ THEN data is of type ‘c’ THEN data is of type ‘e’ Value 7 . 2218.48 AND THEN data is of type ‘f’
Value 6 . 2218.48 AND
Value 1 ≤ 2224.59 AND
Value 3 ≤ 2341.57
THEN data is of type ‘e’
IF Value 1 ≤ 2 349.08 ANDIF Value 2 ≤ 2 698.16 AND IF Value 2 2 . 2349.08 ANDIF Value 8 ≤ 365.89 AND IF Value 1 ≤ 2 500.26
Value 1 . 2500.26 Value 9 ≤ 2212.37 Value 2 ≤ 2341.57 AND Value 4 ≤ 569.77 THEN data is of type ‘f’
THEN data is of type ‘a’ THEN data is of type ‘d’ Value 10 . 2341.57 AND THEN data is of type ‘e’
Value 1 . 2349.08 AND
Value 8 ≤ 2 334.53
THEN data is of type ‘e’
IF Value 8 ≤ 2500.26 AND IF Value 8 ≤ 2212.37 IF Value 1 . 2727.52 AND IF Value 7 ≤ 2 218.48 –
Value 6 ≤ 2 727.5238 THEN data is of type ‘d’ Value 2 . 2349.08 AND THEN data is of type ‘e’
THEN data is of type ‘b’ Value 2 ≤ 2218.48 AND
Value 1 ≤ 2500.26
THEN data is of type ‘e’
IF Value 2 ≤ 2500.26 AND IF Value 3 ≤ 2218.48 IF Value 1 . 2349.08 AND IF Value 8 ≤ 0 AND –
Value 1 ≤ 2619.93 THEN data is of type ‘d’ Value 8 ≤ 20 AND Value 7 . 2218.48 AND
THEN data is of type ‘b’ Value 3 ≤ 2341.57 AND Value 3 . 2218.48 AND
Value 10 . 2174.59 Value 2 . 2218.48
THEN data is of type ‘e’ THEN data is of type ‘e’
be used for the switch open circuit fault case. As the switch larger or smaller batch size of data samples. Too large a
being ‘open’ is a legitimate state of the circuit, the output batch (over 8000) might not reduce false alarms
would show a fault indicator continuously. significantly, whereas too small a batch (below 4000) may
Hence, to detect the introduction of this type of fault, impose additional computational burden. The authors
consecutive fault indications were counted per 0.5 ms worth preferred to use a batch size of 5000 samples as it
of data as opposed to the total number for other fault types conveniently divides the pre-fault and post-fault parts of the
mentioned. This was done based on the notion that since data into two sub-parts each.
the switch open circuit would occur every cycle, only a
large enough consecutive count will suggest a permanent 7 Discussion
switch open circuit failure. Therefore the added code was
modified to count 200 consecutive fault indications per This research stems from the application of power electronic
5000 data samples (i.e. worth 0.5 ms). converters within the zonal shipboard power distribution
This frequency check was found to reduce the number of architecture for US Navy warships and other tightly
false alarms while preserving the speed and reliability of coupled finite inertia systems. As mentioned before, the
fault detection. A similar procedure could be used over a power converter forms an integral part of these topologies
IF Value 3 ≤ 2 3473 AND IF Value 1 . 2942.24 IF Value 1 . 22605.36 IF Value 6 . 23563.6 IF Value 5 . 23630 AND
Value 9 ≤ 22001.3696 THEN data is of type ‘c’ AND AND Value 2 ≤ 2350.48
AND Value 7 ≤ 2 491.18 Value 2 . 21800 AND THEN data is of type ‘a’
Value 1 ≤ 24130.2624 THEN data is of type ‘e’ Value 2 ≤ 2499.73
THEN data is of type ‘a’ THEN data is of type ‘f’
IF Value 1 . 23580.4848 IF Value 1 ≤ 21766.7 AND IF Value 1 . 21766.7 IF Value 1 . 24223.15 –
AND Value 2 ≤ 2 1848.24 AND THEN data is of type ‘e’ AND
Value 2 ≤ 23050.2 Value 9 . 23473 Value 1 ≤ 23654.2 AND
THEN data is of type ‘a’ THEN data is of type ‘d’ Value 2 ≤ 23654.2 AND
Value 3 ≤ 23805.2
THEN data is of type ‘f’
IF Value 2 ≤ 2 4983 IF Value 3 ≤ 23998.16 IF Value 1 ≤ 2 4677.37 IF Value 2 ≤ 21536 AND –
THEN data is of type ‘b’ AND THEN data is of type ‘e’ Value 1 . 23998.16 AND
Value 9 . 23890.93 Value 8 . 23360 AND
THEN data is of type ‘d’ Value 1 ≤ 23654.2
THEN data is of type ‘f’
and the buck converter section was chosen by the authors for 7.2 Rationale for using SCA
analysis into automated fault diagnosis. The research is aimed
at providing high-level diagnostics to aid in decision support The need to understand pertinent failures in a novel system
at system and sub-system levels and focuses on component without any prior benchmarking lead this research into
level and incipient failures to provide a general framework reliability analysis, in turn identifying FMEA as a valid
for fault identification. FMEA enables understanding of starting point. Further, SCA being a known reliability
pertinent component-level failures for the devices which in analysis technique to diagnose design stage circuit faults
turn informs the simulation studies to generate fault data. was studied to check its applicability for the novel
Along with a detailed FMEA, modification of an SCA shipboard distribution architecture.
method for electrical circuits is used to produce usable Thereafter, novel modification of conventional SCA
features that are believed to aid in differentiating operating provided a means to conduct component-level fault
scenarios. signature analysis using a high-level approach with basic
measurements such as input and output currents and
7.1 Proposed diagnostic method compared to voltages. This novel methodology thus gives an analysis
existing research technique stemming from well-known design-stage
reliability analysis which is in turn able to provide
In Section 2, a comprehensive literature review highlights diagnostic indicators. Further, as the explained SCA
research in diagnostics for the related areas of this particular methodology can be derived for other converter circuits, its
research arena. As highlighted in the references, a major application to other types of power electronic converters is
research thread is system reconfiguration, restoration and potentially extendable.
automatic mitigation of faults at the system level. Such
faults are typically ground faults or faults at the main 7.3 Relevance and benefits of proposed approach
motor/generator/prime mover side. Research into the need
for autonomous high-level diagnostics providing decision This research is directed at proposed naval shipboard power
support assistance to the reduced numbers of on-board systems which are envisioned to integrate the plug and play
crew, giving fault information and their risk mitigation for PEBB devices, automated diagnostics and reconfiguration
component-level faults has been ignored. This gap in systems among others. One method is that this integration
research led the authors to conduct studies into component could be facilitated by multi-agent systems [63] that aim to
level diagnostics for the dc – dc power electronic converter provide robustness and extensibility. As mentioned earlier,
since the proposed architectures are envisioned to include the reduction in the numbers of onboard crew with increase
a large number of power electronic sub-systems which as of in automation put added importance on enhanced
now is a relatively new and unproven technology for communication systems with decision support besides the
warships. obvious criticality of accurate fault diagnosis and
The proposed approach beginning at FMEA, feeds monitoring. The nature of such envisioned SPS architecture
further research by providing information about pertinent puts emphasis on condition monitoring and diagnostics at
failures at all levels (FMEA can be conducted at system, various system levels.
sub-system and device/component level), whereas the The diagnostic indicators explained in this paper arise from
modified SCA methodology provides diagnosable features the changes in voltage and current during the various
for faults for the shown converter type. Done together as operating conditions (fault and no-fault). These measured
part of a systematic methodology as proposed, the FMEA quantities change during various operating scenarios to
and modified-SCA approach can be extended to other provide signatures that could be used as diagnostic
power electronic systems to provide high-end diagnosable indicators for those respective cases using a similar rule-
features. It could also potentially provide diagnosable based approach. SCA provides the means of understanding
features for zone- and system-level faults, which forms part circuit connectivity at a device/component level and also on
of future research. a sub-system/zonal level. The novelty proposed here by
822 IET Power Electron., 2012, Vol. 5, Iss. 6, pp. 813 –826
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-pel.2011.0018
www.ietdl.org
modifying conventional SCA provides the potential for recorded (in microseconds) and computing the time needed
single-signal diagnostics for individual inter-connected to diagnose the data (,0.5 ms as shown in Table 2), the
devices. Further, the key to the technique lies in its counting schemed does not have any major detrimental
potential extendibility to more complex systems (a zone in impact on the speed of diagnosis.
the case of notional shipboard systems followed by
interconnected zones), where the circuit connectivity 7.5 Observability condition and effective series
information coupled with a fault diagnostic indicator could resistance
provide valuable additions for a crew’s decision support
which is an important application considering the heavy The experiments reported include four measurement signals,
reduction in crew numbers on future ships. Also, a positive namely the input current, input voltage and output current
is that owing to the approach using one signal for and output voltage. All other internal currents and voltages
diagnostics per device (or per zone), the potential for are computed using well-known circuit laws as explained
adding to the burden of the notional warship’s in Section 5. However, if quantities internal to a circuit
communication network is minimal. are not measurable satisfactorily by indirect methods, then
Ongoing research aims to investigate the novel modified additional sensors may be needed to obtain these
SCA methodology reported in this paper to generate single- measurements.
signal fault diagnostic indicators at various levels of the For low-voltage supplies, capacitor ESR can be the greatest
system’s power distribution architecture. This paper reports contributor to voltage ripple in comparison with the voltage
an application of the methodology at one such level, that is, ripple across the capacitive element of the capacitor. In the
the device level (buck converter). Further research is case studied in this paper, the voltage supply is high and
actively aimed at extending this methodology to higher the relative contribution to voltage ripple from ESR is
levels of the network wherein not only device-level small. In general, the technique is capable of including ESR
diagnostics is available but ideally ship-wide diagnostics in the algorithm and derivation process. The main change is
with resourceful decision support can be provided. in the modified-SCA that is performed, wherein ESR could
be considered as a resistor in series with the capacitor. This
7.4 Fault introduction procedure in hardware added ‘component’ changes the determinant obtained from
circuit and avoiding false alarms the modified SCA method, although the analysis and
derivation technique to obtain the diagnostic feature is
A hardware setup was used to simulate specific faults and fundamentally same as proposed.
failures that are known to occur in the semiconductor
devices (typically open-circuit failures which usually have a 7.6 Scalability issue of proposed method
pre-cursor of a rapid short-circuit time prior to bond wires
arcing in plastic packages). In the passive components the The advantage of utilising data from local measurements such
authors simulated the effects of electrolyte leakage in as input/output measured currents and voltages as well as
capacitors (an incipient fault) or rapid turn – turn failures in known circuit parameters avoids the use of additional
inductor windings. These tests are realistic in terms of the sensors. Furthermore, the novel diagnosis method presented
overall impact on the performance of the converter circuit. makes use of this local data for each converter to produce
The halving of net capacitance by parallel connected usable features from the determinant of the modified
capacitors (and halving of net inductance by series generalised connection matrix for the respective device.
connected inductors) is used to simulate respective fault This is irrespective of the number of interconnected devices
cases. Although the step changes are not representative of as each converter will have its own matrix. Thus, for a
practical failure modes, it provides an accelerated test scaled up complex system with many power electronic
method in order to assess the suitability of the proposed converters, the modified matrix would be constructed
technique to detect and diagnose such faults. However, it is separately for each device to provide localised diagnosis at
acknowledged that the actual failure because of such the component level for each converter. This concept
incipient faults is a slower process potentially occurring prevents the modified generalised connection matrix from
over several hundred hours of operation. An alternative becoming too large to pose any issues relating to
method for simulation of a capacitor fault of this type could computational burden.
be to employ differently aged capacitors in parallel, such
that the progressive blocking of individual capacitors at 7.7 Future research plans
designated instants in time would switch to the more ‘aged’
or worn-out capacitor with prior-estimated (or known) In order to form a topology-independent approach, the
changes in related parameters [64]. This method potentially proposed methodology is to be applied to dc – dc converter
may provide more accurate capacitor fault data and is circuits such as the boost and buck – boost converters. The
anticipated to be applied in future research plans. verification could be similarly carried out using hardware
For the switch open and short circuit faults, the faults setups. The next step is to implement the methodology on a
are introduced when the switch is open and closed, real-time basis, for example, using field programmable gate
respectively. However, in a real system it is difficult to arrays [65]. Fig. 13 shows the basic setup to investigate
predict the switch’s state when the passive component fails real-time diagnostics.
because of electrolyte leakage or inductor windings fault. This research was conducted on a single isolated
In this case, however, as fault introduction is manual, it is buck converter circuit with fixed parameters. Applying
known that the faults for both the capacitor and inductor the proposed diagnosis technique in a grid-connected
have been introduced when the switch S1 is open (off state). environment is a matter of further investigation. This could
The counting scheme explained in Section 6.6 checks for ideally be done by using the converter with loads (such as
consistency in the data stream thus reducing the possibility motors) the parameters of which can be changed when
of a false diagnosis. Owing to the time scales of the data faults are introduced in the converter circuit. The feasibility
824 IET Power Electron., 2012, Vol. 5, Iss. 6, pp. 813 –826
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-pel.2011.0018
www.ietdl.org
method on a real-time basis for a buck converter hardware 20 Butler-Purry, K.L.: ‘Multi-agent technology for self-healing shipboard
circuit. In addition to component level fault diagnostics, the power systems’. Proc. 13th Int. Conf. on Intelligent Systems
Application to Power Systems, 6–10 November 2005, p. 5
applicability of the proposed method for sub-system (zone) 21 Butler-Purry, K., Srivastava, S.K.: ‘Expert system-based reconfiguration
and system-wide faults (e.g. ground faults) is also an aim of shipboard power distribution systems’. 2008 IEEE Power and Energy
for further study. Society General Meeting – Conversion and Delivery of Electrical
Energy in the 21st Century, 20–24 July 2008, pp. 1 –5
22 Butler-Purry, K.L., Sarma, N.D.R.: ‘Self-healing reconfiguration for
restoration of naval shipboard power systems’, IEEE Trans. Power
9 Acknowledgments Syst., 2004, 19, (2), pp. 754– 762
23 Baran, M.E., Mahajan, N.: ‘System reconfiguration on shipboard DC
The authors thank the US Office of Naval Research for zonal electrical system’. IEEE Electric Ship Technologies Symp.,
funding this research. They also authors thank the support 25– 27 July 2005, pp. 86– 92
of David Cartes, Karl Schoder and Sanjeev Srivastava of 24 Rodriguez, J., Lai, J.-S., Zheng Peng, F.: ‘Multilevel inverters: a survey
of topologies, controls, and applications’, IEEE Trans. Ind. Electron.,
the Centre for Advanced Power Systems at Florida State 2002, 49, (4), pp. 724–738
University for their support. 25 Sleszynski, W., Nieznanski, J., Cichowski, A.: ‘Open-transistor fault
diagnostics in voltage-source inverters by analyzing the load currents’,
IEEE Trans. Ind. Electron., 2009, 56, (11), pp. 4681– 4688
10 References 26 Kim, T.-J., Lee, W.-C., Hyun, D.-S.: ‘Detection method for open-circuit
fault in neutral-point-clamped inverter systems’, IEEE Trans. Ind.
1 Soman, R.R., Davidson, E.M., McArthur, S.D.J.: ‘Using functional Electron., 2009, 56, (7), pp. 2754–2763
failure mode and effects analysis to design the monitoring and 27 Lezana, P., Aguilera, R., Rodriguez, J.: ‘Fault detection on multicell
diagnostics architecture for the zonal MVDC shipboard power converter based on output voltage frequency analysis’, IEEE Trans.
system’. IEEE Electric Ship Technologies Symp., (ESTS 2009), Ind. Electron., 2009, 56, (6), pp. 2275–2283
20– 22 April 2009, pp. 123–128 28 Young Kim, S., Nam, K., Song, H.-K., Kim, H.-G.: ‘Fault diagnosis of a
2 Hegner, H., Desai, B.: ‘Integrated fight through power’. IEEE Power ZVS DC– DC converter based on DC-link current pulse shapes’, IEEE
Engineering Society Summer Meeting, July 2002, vol. 1, pp. 336 –339 Trans. Ind. Electron., 2008, 55, (3), pp. 1491–1494
3 Amy, J.V.: ‘Considerations in the design of naval electric power 29 Zidani, F., Diallo, D., Benbouzid El Hachemi, M., Nait-Said, R.: ‘A
systems’. IEEE Power Engineering Society Summer Meeting, 25–25 fuzzy-based approach for the diagnosis of fault modes in a voltage-fed
July 2002, vol. 1, pp. 331– 335 PWM inverter induction motor drive’, IEEE Tran. Ind. Electron.,
4 Mair, A.J., Soman, R.R., Davidson, E.M., et al.: ‘Intelligent distributed 2008, 55, (2), pp. 586–593
control for shipboard power systems based on fault diagnosis and 30 Khomfoi, S., Tolbert, L.M.: ‘Fault diagnosis and reconfiguration for
condition monitoring’. Engine as a Weapon III, Portsmouth, UK, multilevel inverter drive using AI-based techniques’, IEEE Trans. Ind.
23– 24 June 2009 Electron., 2007, 54, (6), pp. 2954–2968
5 Mikulak, R.J., McDermott, R., Beauregard, M.: ‘The basics of FMEA’ 31 Pan, Y., Silveira, P.M., Steurer, M., Baldwin, T.L., Ribeiro, P.F.: ‘A fault
(CRC Press, 2nd edn.), Dec. 2008 location approach for high-impedance grounded DC shipboard power
6 Ericsen, T., Hingorani, N., Khersonsky, Y.: ‘Power electronics and distribution systems’. IEEE Power and Energy Society General
future marine electrical systems’, IEEE Trans. Ind. Appl., 2006, 42, Meeting – Conversion and Delivery of Electrical Energy in the 21st
(1), pp. 155– 163 Century, 20–24 July 2008, pp. 1 –6
7 Logan, K.P.: ‘Intelligent diagnostic requirements of future all 32 Charfi, F., Lesecq, S., Sellami, F.: ‘Fault diagnosis using SWT and
electric ship integrated power systems’, IEEE Trans. Ind. Appl., 2007, Neyman Pearson detection tests’. IEEE Int. Symp. on Diagnostics for
43, (1), pp. 139–149 Electric Machines, Power Electronics and Drives, SDEMPED 2009,
8 RTDS Technologies: ‘Real-time digital simulator controls library 31 August, 6–3 September 2009, pp. 1– 6
manual’ (2002) 33 Tsoumas, I.P., Safacas, A.N.: ‘Investigation of a novel approach for fault
9 Real Time Digital Simulator Tutorial Manual, RSCAD version, diagnosis in power electronic converters: the case of the subsynchronous
September 2005 cascade drive’. IEEE Int. Symp. on Diagnostics for Electric Machines,
10 Rodriguez, J., Pontt, J., Alzarnora, G., Becker, N., Einenkel, O., Weinstein, Power Electronics and Drives, 2007, SDEMPED 2007, 6– 8
A.: ‘Novel 20-MW downhill conveyor system using three- September 2007, pp. 58–63
level converters’, IEEE Trans. Ind. Electron., 2002, 49, (5), pp. 1093–1100 34 Wang, F., Lai, R., Yuan, X., Luo, F., Burgos, R., Boroyevich, D.:
11 Rodriguez, J., Bernet, S., Wu, B., Pontt, J.O., Kouro, S.: ‘Fault detection and protection of three-level neutral-point-
‘Multilevel voltage-source-converter topologies for industrial medium- clamped PWM voltage source converters’. IEEE Industry
voltage drives’, IEEE Trans. Ind. Electron., 2007, 54, (6), Applications Society Annual Meeting, (IAS ’08), 5 –9 October 2008,
pp. 2930– 2945 pp. 1– 7
12 Cheng, Y., Qian, C., Crow, M.L., Pekarek, S., Atcitty, S.: ‘A comparison 35 Wang, F., Lai, R., Yuan, X., Luo, F., Burgos, R., Boroyevich, D.:
of diode-clamped and cascaded multilevel converters for a STATCOM ‘Failure-mode analysis and protection of three-level neutral-point-
with energy storage’, IEEE Trans. Ind. Electron., 2006, 53, (5), clamped PWM voltage source converters’, IEEE Trans. Ind. Appl.,
pp. 1512– 1521 2010, 46, (2), pp. 866–874
13 Alepuz, S., Busquets-Monge, S., Bordonau, J., Gago, J., Gonzalez, D., 36 Bose, B.: ‘Power electronics – an emerging technology’, IEEE Trans.
Balcells, J.: ‘Interfacing renewable energy sources to the utility grid Ind. Electron., 1989, 36, (3), pp. 403–412
using a three-level inverter’, IEEE Trans. Ind. Electron., 2006, 53, (5), 37 Rothenhagen, K., Fuchs, F.W.: ‘Performance of diagnosis methods for
pp. 1504– 1511 IGBT open circuit faults in three phase voltage source inverters for
14 Veerachary, M.: ‘Analysis of fourth-order DC– DC converters: a flow AC variable speed drives’. European Conf. on Power Electronics and
graph approach’, IEEE Trans. Ind. Electron., 2008, 55, (1), pp. 133 –141 Applications, 2005, p. 10
15 Barrena, J.A., Marroyo, L., Vidal, M.A.R., Apraiz, J.R.T.: ‘Individual 38 Lu, B., Sharma, S.: ‘A literature review of IGBT fault diagnostic
voltage balancing strategy for PWM cascaded H-bridge converter- and protection methods for power inverters’. IEEE Industry
based STATCOM’, IEEE Trans. Ind. Electron., 2008, 55, (1), Applications Society Annual Meeting, (IAS ’08), 5 –9 October 2008,
pp. 21–29 pp. 1– 8
16 Baran, M.E., Mahajan, N.R.: ‘DC distribution for industrial systems: 39 Chen, Y.-M., Wu, H.-C., Chou, M.-W., Lee, K.-Y.: ‘Online failure
opportunities and challenges’, IEEE Trans. Ind. Electroni, 2003, 39, prediction of the electrolytic capacitor for LC filter of switching-mode
(6), pp. 1596– 1601 power converters’, IEEE Trans. Ind. Electron., 2008, 55, (1),
17 Ciezki, J.G., Ashton, R.W.: ‘Selection and stability issues associated pp. 400–406
with a navy shipboard DC zonal electric distribution system’, IEEE 40 Imam, A.M., Divan, D.M., Harley, R.G., Habetler, T.G.: ‘Real-time
Trans. Power Deliv., 2000, 15, (2), pp. 665– 669 condition monitoring of the electrolytic capacitors for power
18 Momoh, J.A., Kaddah, S.S., Salawu, W.: ‘Security assessment of DC electronics applications’. 22nd Annual IEEE Applied Power
zonal naval-ship power system’. Large Engineering System Conf. on Electronics Conf. (APEC 2007), 25 February–1 March 2007,
Power Engineering LESCOPE ’01, 2001, pp. 206– 212 pp. 1057– 1061
19 Momoh, J.A., Ishola-Salawu, A.S.: ‘A new arcing fault modeling and 41 Amaral, A.M.R., Cardoso, A.J.M.: ‘Use of ESR to predict failure of
detection technique for navy IPS power system’. IEEE Power output filtering capacitors in boost converters’. IEEE Int. Symp. on
Engineering Society General Meeting, 2006, p. 7 Industrial Electronics, 4 –7 May 2004, pp. 1309– 1314
826 IET Power Electron., 2012, Vol. 5, Iss. 6, pp. 813 –826
& The Institution of Engineering and Technology 2012 doi: 10.1049/iet-pel.2011.0018