09-02-2013, 11:13 AM
Health Monitoring, Fault Diagnosis and Failure Prognosis Techniques for Brushless Permanent Magnet Machines
1Health Monitoring.pdf (Size: 361.29 KB / Downloads: 339)
Abstract
Over the past few years, many researchers have been
attracted by the challenges of electrical machines’ fault diagnosis
and condition monitoring, which provide early warnings that could
help schedule necessary maintenance to avoid catastrophic
consequence. With advancements in the use of rare-earth magnets,
Brushless Permanent Magnet Machines are widely used in industry
recently, which has led to the development of numerous fault
diagnosis techniques. Considerable papers have presented reviews
and compared condition monitoring and fault diagnosis methods for
induction machines, but none for Brushless Permanent Magnet
Machines. To make a difference, this paper presents a
comprehensive survey of modern research advancements and stateof-
the-art in health monitoring, fault diagnosis and prognosis
techniques for Brushless Permanent Magnet Machines. The
symptoms of each type of fault and the principles of diagnosis
process are also described and discussed.
INTRODUCTION
Over the past decade, Permanent Magnet Synchronous
Machines (PMSMs) and Brushless DC machines (BLDCs)
have gained significant popularity in the industry, especially
where high performance is required, owing to higher
efficiency, high output power to volume ratio, high torque to
current ratio, etc. Some of the commonly occurring faults in
PMSMs are eccentricity, bearing failure, demagnetization of
permanent magnets, short circuit in the stator or armature
winding, etc. Health monitoring and fault diagnosis of the
machine could help in scheduling preventive maintenance to
length their lifespan and avoid catastrophic system failure.
Basic steps for a machine diagnosis scheme are illustrated in
Fig.1, where dashed lines indicate non-necessary steps.
Considerable papers have presented reviews and compared
condition monitoring and fault diagnosis methods for
induction machines [1]-[4], but none for Brushless Permanent
Magnet Machines.
TYPES OF FAULTS IN PM MACHINES
In an electric machine, faults can occur in the rotor/field,
stator/armature, inverter, or mechanical components
connected to it. This paper discusses a permanent magnet
machine without focusing on associated inverter faults and
bearing faults. Fig. 2 illustrates the most frequently
encountered problems for electric machines [5]. Their causes
and symptoms are presented in this section.
Permanent Magnetic Faults
For permanent magnet machines, field fault typically refers
to a failure in the permanent magnets, where demagnetization
is the most common issue. The demagnetization could be
uniform over all poles or partial over certain region or poles.
Conditions that could cause permanent magnets in a PMSM
to demagnetize include
1) High operation temperature/Cooling system malfunction
2) Aging of magnets
3) Corrosion of magnets
4) Inappropriate armature current
FAULT SIGNATURE DETECTION TECHNIQUES
Frequency Signature Analysis
Stator current frequency analysis is the most frequently
used technique for machine fault diagnosis. It is usually
called Motor Current Signature Analysis (MCSA). Fast
Fourier transform (FFT) method is widely used for frequency
analysis. Some specific harmonics in the stator winding
current spectrum can be detected as a signature of specific
fault. Since only current measurement is required, this
method can be used for simultaneous multi-fault detection. In
addition, this technique is also non-invasive, and costeffective.
In dynamic eccentricity case, frequency components in the
order of 1/P exist in the inductance versus position profile,
where P is the number of pole pairs. Since the position of
minimum air gap rotates, harmonics can be found in the stator
current spectrum.
FAULT IDENTIFICATION TECHNIQUES
Sometime, extracted fault features are not easily
distinguished from fault-free cases or other fault types,
especially when disturbances and noise are not negligible.
Therefore, several researchers have reported artificial
intelligence (AI) algorithms for faults isolation. This
approach maps extracted features to specific fault categories.
Some frequently used fault identification tools are: artificial
neural network (ANN), fuzzy logic, expertise system, support
vector machine, etc.
ANN mimics the structure of human brain, which consists
of numerous processing unit-neurons, to form a complex
adaptive network to perform multi-input/multi-output
mapping, even though every neuron has a very simple
function. Usually, complex systems will have many hidden
layers or feedbacks to enhance the ANN performance.
However, the main challenges are that such a system often
requires massive computational device and large storage
memory. In real application, commonly used ANNs in
literatures are three layers feed forward structure in order to
simplify the implementation and lower the hardware
requirements.
CONCLUSION
A brief review of armature, field and mechanical faults
occurs to PMSM and their diagnosis schemes have been
presented in this paper. So far spectrum signature analysis
based methods and machine model based methods are the
mostly preferred. Apart from Fourier Transform based
techniques, other spectrum manipulation algorithms, such as
higher order spectrum, wavelet transforms show better
performance for various speed conditions. Artificial
intelligence algorithms are also discussed in this paper, which
is necessary to distinguish the fault signatures.