Detection and Separation of Faults in Permanent Magnet Synchronous Machines using Hybrid Fault-Signatures
Item Type Conference Paper
Authors Ullah, Zia;Im, JunHyuk;Ahmed, Shehab
Citation Ullah, Z., Im, J., & Ahmed, S. (2022). Detection and Separation of Faults in Permanent Magnet Synchronous Machines
using Hybrid Fault-Signatures. 2022 IEEE Energy Conversion Congress and Exposition (ECCE). https://doi.org/10.1109/
ecce50734.2022.9947448 Eprint version Post-print
DOI 10.1109/ECCE50734.2022.9947448
Publisher IEEE
Rights This is an accepted manuscript version of a paper before final publisher editing and formatting. Archived with thanks to IEEE.
Download date 2024-01-08 20:07:07
Link to Item http://hdl.handle.net/10754/686081
XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE
Detection and Separation of Faults in Permanent Magnet Synchronous Machines using Hybrid Fault-
Signatures
Zia Ullah
Department of Computer, Electrical and Mathematical Science and Engineering King Abdullah University of Science and
Technology (KAUST) Thuwal, Saudi Arabia [email protected]
JunHyuk Im
Department of Electrical Engineering Incheon National University
Incheon, Republic of Korea [email protected]
Shehab Ahmed
Departmernt of Computer, Electrical and Mathematical Science and Engineering King Abdullah University of Science and
Technology (KAUST) Thuwal, Saudi Arabia [email protected]
Abstract— As digitalization in electric motors accelerates, online condition monitoring systems are becoming more popular, allowing unplanned downtime to be detected at its initial stage. Individual faults in motors are effectively diagnosed. However, due to identical signatures, fault separation and/or identification remain a challenge. This study presents the detection and separation of inter-turn short, demagnetization, static eccentricity, bearing, and the combination of these faults in permanent magnet synchronous machines. Hybrid fault signatures are used by monitoring the frequency spectrum of stator current, vibration, and induced voltage in the airgap. A planer-shaped airgap search coil (PASC) is employed to extract the induced voltage of each stator tooth. Faults-related anomalies in the induced-voltage, vibration, and the stator current caused are monitored. Any deviation in either signal is considered as generic fault indicator. Furthermore, specific fault features in all signals are used to classify these faults with improved accuracy. The PASC used in this study can also identify the location of the inter-turn short fault and the severity of demagnetization fault.
The proposed method is verified using the finite element method simulation and experiments.
Keywords—inter turn fault, demagnetization fault, eccentricity fault, bearing fault, fault diagnosis, PMSM.
I. INTRODUCTION
Condition monitoring of electric motors and generators has grown in the industrial, automotive, aerospace, and renewable energy markets to avoid unexpected downtime due to several potential faults. Equipment failures are estimated to account for approximately 30% of downtime in manufacturing plants [1]. Online condition monitoring solutions are becoming more prevalent as digitalization accelerates. In the online monitoring technique, critical equipment is constantly monitored by data based software.
Any fault that occurs are communicated to the maintenance crew months in advance. As a result, the fault is ended before it spreads. This is owing to enhanced technology reliability and availability, which has provided buyers with a competitive advantage. In this regard, online condition monitoring systems are an excellent way to minimize unplanned faults to zero [2].
Permanent magnet synchronous machines (PMSMs) are
considered the best candidates for the aforementioned application areas due to their high torque and power density, compact structure, and precise control [3]. Despite their attractive features, PMSMs are prone to various faults such as winding inter-turn short fault (ITSF), irreversible demagnetization fault (IDF), eccentricity fault (EF), and bearing fault (BF) are common in PMSMs [4].
Consequently, machine health monitoring remains an integral part of the motor drive system and it is gaining more popularity owing to the growing number of safety-critical applications. An efficient condition monitoring system should be accurate, reliable, and cost-effective. Despite the existing research, the diagnosis of faults in different PMSMs is still an open problem [5]. In fact, most of the fault diagnosis methods have one or more insufficiency from the economical or technical point of view. Accuracy is the main limitation in the currently used methods for the PMSMs quality assessment.
Several model-based, signal-based, and knowledge-based (data-driven) techniques for fault diagnosis has been proposed so far [6], [7]. Among the various fault diagnosis techniques, the motor current and voltage signature analysis [8], [9], search coil [10], winding impedance [11], vibration [12], torque ripple [13], [14], and load angle based technique are commonly discussed [15]. MCSA and MVSA are the most researched methods and considered effective noninvasive techniques for some faults in certain operating conditions. All these methods have their pros and cons, some of them are performing well in specific operating conditions.
For instance, MCSA performs well in the detection of ITSF in a steady-state condition. Similarly, search coil-based techniques are good in IDF detection. With some limits, all these techniques correctly diagnose the faults for which they were developed. The common drawback of all available approaches is that they cannot detect more than one or sometimes two types of faults. Furthermore, they are unable to distinguish between distinct types of faults.
Some fault separation techniques have already been proposed for various motor faults. J. Hong et al. used the d- axis inductance curve as a signature to detect and separate uniform IDF and eccentricity fault [16]. In this technique a signal is injected in the d-axis direction to measure the equivalent d-axis “differential” inductance. However, this
method can only be implemented in offline condition and the motor needs to disassemble. Similarly, Haddad et al.
measured incremental d-axis inductance and used k-Nearest Neighbor classifier for the separation of SE and ITSF [17].
The MCSA and MVSA are used to identify between dynamic eccentricity, static eccentricity, and IDF for small PMSMs. However, the accuracy of this method affects when the core saturates. The aforementioned techniques utilize different motor input/output signals to diagnose and classify various commonly occurred faults. However, detection and separation of different frequently occurring faults cannot be detected and separate using one type of signature.
Furthermore, it is extremely challenging to separate more than two types of faults by using simple signal analysis like MCSA or MVSA. Consequently, numerous machine and deep learning-based techniques are employed for accurate classification of faults [18]. Principal component analysis and support vector machine are used to classify faults using current residuals in PMSM [19]. Wavelet packet along with 1-D convolutional neural network was used to classify three types of faults in PMSM [20]. These methods are computationally complex and there is no concreate method presented so far which can classify these faults in real-time.
Nevertheless, a single method that combines the best fault signatures for multiple faults detection and classification remains necessary. An optimal diagnosis technique should detect and differentiate between all frequently occurred faults in the motor drive system.
The contribution of this paper is using hybrid fault signatures by combining the stator current, vibration, and induced voltage in a single algorithm to detect and classify ITSF, IDF, EF, BF and the combination of these faults. A new planer-shaped airgap search coil (PASC) is used for obtaining induced voltage. PASC can also be used for position estimation of the machine [21]. The PASC and current signal can directly detect and distinguish between ITSF, IDF and SE in online condition. To improve the accuracy vibration signal are used to further verify the occurrence of these faults. Furthermore, vibration signal helps in diagnosis of BF and mix faults. The data produced from PASC and the stator current complement each other in accurately detecting and separating faults.
II. MAJOR FAULTS IN PMSMS
Several types of electrical and mechanical faults occur in the PMSMs which severely affect the reliability of the machine and cause unexpected downtime. The most commonly occurring faults in PMSMs includes ITSF, IDF, EF, and bearing faults. These four faults’ accounts for over 70% of total faults in electric motors.
A. Inter Turn Short Fault (ITSF)
ITSF is one of the most frequently occurring faults in induction and PM type machines. When two or more turns in the same phase short circuited due to insulation degradation results in ITSF. This fault spread very rapidly and can cause the sudden shutdown of the entire system. After the occurrence of ITSF, the faulty winding turns act as an extra circuit connected to flux linkages in the rest of healthy phases shown in Fig. 1 (a). Thus, a high circulating current
(a) (b)
Fig. 1. ITSF. (a) Schematic diagram, (b) Experimental result of the phase and circulating current under 8.33% ITSF in the benchmark IPMSM.
is induced in the faulty winding turns [22]. The circulating current is four to five times larger than the phase current and flows in the opposite direction as shown in Fig. 1(b). The heat produced is proportional to the square of the circulating current. Under ITSF, the losses in the motor increase significantly because the motor draw higher current for the same amount of load compared to healthy condition. This fault spread very rapidly and if not prevented it can damage the whole winding within ten minutes. Furthermore, due to the extra heat the PMs can also be damaged.
B. Irreversible Demagneiztion Fault (IDF)
Permanent reduction in the strength (residual magnetic flux density) of a PM is called IDF. Various factors such as aging, physical damage, high operating temperature, and severe field weakening can cause IDF. Once the residual magnetic flux density drops below the knee-point, then it is irrecoverable. IDF causes various variations in the magnetic flux distribution of the machine which further severely affects the behavior of the IPMSM. These variations reflect in the machine linkage flux, which can be determined by the
(a)
0 30 60 90
120 150
180
210 240
270
300 330
0.5T
Healthy 1.0T IDF
(b)
Fig. 2. Simulation results of IDF. (a) different severities of IDF, (b) magnetic flux density distribution under healhty and IDF condition.
input voltage and current. Fig. 3(a) and (b) shows different possibilities of partial IDF designed in FEM and the simulation result of the air-gap magnetic flux density distribution of the IPMSM under healthy and partially IDF condition. The reduction in magnetic flux density due to IDF is evident.
C. Eccentricity Fault (EF)
Eccentricity fault in the stator and rotor mainly occurs due to mechanical reasons. This fault is categorized as static eccentricity fault, dynamic eccentricity fault, and hybrid eccentricity. This study focused on static eccentricity fault.
In case of static EF, the symmetrical axis of the rotor is concentric with the rotor axis. However, they are dislocated with respect to stator symmetrical axis as shown in Fig. 4.
The self- and mutual inductances in rotor phases as well as the mutual inductances across the rotor and stator, are function of the angular position of the rotor just like in healthy condition. In normal condition, the rotor is centered at the stator bore and has identical airgap between the stator and rotor. Subsequently, balanced magnetic forces are generated in the opposite directions, but when the static EF emerges, the airgap lessens (g2) on one side, whereas it gets bigger on the other side (g1) at any gyration angle which results in bigger absorption force through the shorter gap and lower force at bigger airgap. Fig. 5 shows the magnetic flux density of an IPMSM under healthy and 80% static eccentricity fault. The variation in magnetic flux density due to smaller and bigger airgap can be clearly seen.
D. Bearing Fault
Bearing fault is the most frequently occur fault in the electric motor which accounts for above 40% among all types of faults. Generic deep-groove ball bearing consists of the outer race, inner race, and balls, as shown in Fig. 6(a).
Lubricant is applied to the rolling elements of the bearing.
Every bearing has a defined life cycle. Apart from mechanical causes such as overloading or other physical damage electrical stresses also play a vital role in bearing degradation. In inverter-controlled machines, the common- mode voltages and the parasitic capacitance led to bearing current [23]. This bearing current produce heat which leads to melting the lubricant and start affecting the surface of both races and the ball of the bearing. Normally, the outer system, the gradual degradation of bearing due to electrical stress can still lead to BF, which needs to be detected at its early stage to avoid further damage.
Fig. 3. IPMSM with static eccentricity fault along x-axis.
0 60 120 180 240 300 360
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Magnetic flux density [T]
Mechanical angle
Healthy, Static EF g2: Minimum airgap
g1: Maximum airgap
Fig. 4. FEM result of magnetic flux density deistribution of the benchmark IPMSM under healthy and static EF conditions.
race of the bearing is the most vulnerable to electrical stress and in majority of cases the outer race of the bearing damage first. The major sign of the bearing fault is the huge vibration and acoustic noise due to the rough surface of the bearing. The magnitude and pattern of the vibration depend upon the severity and damaged portion of the bearing. Fig.
6(b) shows the vibration signal of the bearing with outer race damaged. Machines are carefully designed and operated to avoid BF. However, even in a very vigorous
(a)
(b)
Fig. 5. (a) Schematic of deep groove bearing, (b) vibration signal of BF.
III. FAULTS DIAGNOSIS IN PMSMS A. MCSA based Faults Signatures
The most prevalent technique for diagnosing electrical faults in motors is motor current signature analysis (MCSA).
FFT analysis of the frequency spectrum of the stator current is still commonly utilized as a fundamental diagnostic tool.
This method's appeal stems from its simplicity, low cost, and online machine condition monitoring capabilities. The frequency spectrum of an IPMSM is investigated in depth in this paper under healthy, ITSF, IDF, EF, BF, and mix fault conditions. Experiments were performed on an IPMSM
having fractional slots and concentrated winding. The detailed parameters of the machine are given in Table I. As shown in Fig. 7, the FFTs of stator current were collected up to 2kHz bandwidth when the benchmark IPMSM was operating at 3000 rpm (150Hz) and 60% load. The windings in the motor under test are concentrated in fractional slots.
In the normal or healthy case, the most dominant harmonics are the 5th, 7th, 11th, and 13th. When ITSF has been applied by shortening three winding turns (4.01% ITSF) in Phase-A. The symmetry of the magnetic flux distribution is disrupted due to reduction in the healthy turns in phase-A and the reverse magnetic flux caused by the shorted turns.
As a result, a significant 3rd and 9th harmonic appeared in the stator current as highlighted in Fig. 7. The second order harmonic can also be seen although its magnitude is very small, and inconsistent. The magnitude of the 3rd and 9th harmonics is significant in this case, although this can vary depending on the severity of the fault. These harmonics can be used as fault indicators of ITSF.
In next case, IDF was applied by replacing small PMs in two poles of rotor. Partial IDF was demonstrated by the reduction in the strength of a few magnet poles in the rotor.
In case of partial IDF, the magnetic field produced by rotor is no longer uniform, and this phenomenon is reflected in the currents and voltages of the motor. The IDF caused a significant 2nd and 4th order harmonics as shown in Fig. 7.
Furthermore, it significantly reduced the magnitude of the 5th harmonic which were larger in case of a healthy condition. The presence of 2nd and 4th order harmonics, as well as a decrease in 5th order harmonic, can be employed as IDF indicators under steady-state conditions. MCSA is the simplest and most cost- effective approach for fault diagnosis. However, these harmonics are highly sensitive to noise and especially the bandwidth of current and speed controllers in the field-oriented control drive system. In general, if the speed controller's bandwidth is low, the voltages will show the typical harmonics of a certain faults.
They will, however, be present in currents if the bandwidth is high. Current controllers can distort this effect if their bandwidth is low; the defect is obvious in currents and, if high, in voltages. In other words, because the control system produces clear sinusoidal current signals, greater harmonics must be present in voltages. Furthermore, the considered fault signatures such as 3rd harmonic can also be caused by mix fault such as BF plus EF fault as shown in Fig. 7.
Moreover, in eccentricity and bearing faults there is no Table I. Parameter of the IPMSM used in this study
Parameter Unit Values
Rated power Watt 400
Rated speed rpm 3500
Rated current Arms 10.32
Rated torque Nm 1.1
Ld mH 0.92
Lq mH 1.35
Phase resistance ohm 0.07
Poles - 6
DC bus voltage Volt 60
Turns per Phase - 72
PM type - NdFeB
Slots - 9
0 200 400 600 800 1000 1200 1400 1600 1800 2000
-60 -40 -20 0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
-60 -40 -20 0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
-60 -40 -20 0 20
0 200 400 600 800 1000 1200 1400 1600 1800 2000
-60 -40 -20 0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
-60 -40 -20 0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
-60 -40 -20 0
Current (dB)
Healthy f=150Hz
5f 7f
11f 13f
Current (dB)
f=150Hz ITSF
11f 13f 5f 7f
3f
2f 9f
Current (dB)
f=150Hz IDF
11f 13f 5f 7f
4f 2f
10f 0.6f
Current (dB)
Eccentricity f=150Hz
13f 11f 7f
5f 3f 2f
Current (dB)
Bearing Fault f=150Hz
11f 13f 5f 7f
3f 2f
Current (dB)
Frequency (Hz)
Bearing + Eccentricity Mix f=150Hz
11f 13f 5f 7f
2f 3f
Fig. 6. Experimental results of the frequency specturm of the stator current under healthy, ITSF, IDF, Eccentricity, BF, and mix fault.
significant additional harmonics that can be used for the detection of these faults. Hence, MCSA is not the sufficient fault indicator for all four types of fault classification with improved accuracy.
B. PASC based Faults Signatures
The search coil has been used successfully for fault diagnostics. The classic search coil, on the other hand, cannot be placed in all commercially available motors.
Furthermore, it necessitates additional room in the slot. A planer-shaped airgap search coil (PASC) made of flexible PCB is employed in this work. PASC is easy to install in any machine by just inserting it into the airgap.
In PMSM, the mechanical torque of the motor is generated by airgap flux [18]. The airgap flux is the combination of the stator flux and rotor magnet flux given in (1). Where Φ airgap is airgap flux, Φs is armature reaction
flux, and Φr is rotor magnetic flux. As PASC is installed in airgap, Φ airgap is magnetically linked to PASC. The linkage-flux λp can be expressed by (2).
Φairgap= Φs + Φr (1) λp = Np ∙ Φr + Mp ∙ Is (2) Where Np is the number of turns of PAFS, Mp is mutual inductance between stator winding and PAFS, and Is is stator current. Thus, the BEMF considering the design of PASC and airgap flux can be derived. From derivative of (3), the total BEMF induced to PASC can be expressed by (3).
Where Ep is the induced voltage of PASC. Equation (3) can be rewritten as (4).
𝑑
𝑑𝑡𝜆𝑝 = 𝐸𝑝 (3) 𝐸𝑝= 𝑀𝑝∙𝑑
𝑑𝑡𝐼𝑠 + 𝑁𝑝∙ 𝑑
𝑑𝑡Ф𝑟 (4) The induced voltage Ep can be easily obtained and analyzed during the operation of the machine. The PASC is designed in ANSYS Maxwell. Fig. x depicts the geometric parameters and schematic design of the PASC. The minimum wire spacing, and wire width are both 0.1mm as per manufacturing availability. The total thickness of this flexible PCB is 0.104mm. Given that the airgap length of a common motor in the field is 0.3mm or greater, this thickness is tolerable. The number of turns in PASC is selected as 37, total thickness 0.104 mm, series resistance 70ohm, series inductance 35 µH, and BEMF constant 0.016.
The flexible PCB is thin enough to fit within the airgap.
Furthermore, the permeability of flexible PCB is close to that of air, the airgap flux distortion caused by the coil is minimal. The materials used are copper and polyimide, with relative permeability of 0.99996 and 1, respectively. The power dissipation by PASC is low due to the high impedance of the ADC that is coupled to the PASC for output signal detection. As a result, attaching PASC has no effect on the motor's performance. Fig. 8(a) depicts the manufactured PASC, whereas Fig. 8(b) depicts three mounted PASC on the pole shoes of the stator of the benchmark IPMSM.
Fig. 9 shows the experimental result of the induced voltage at different speeds obtained by PASC when the motor was rotated in generating mode. A very clear signal from low to high speed with consistent nature was obtained that validate the design of the PASC. The induced voltage of the PASC in motoring condition when the IPMSM was rotating at 3000 rpm and 60% load is shown in Fig. 10. In healthy condition, the induced voltage signal has a symmetrical shape, Fig. 10(a). However, in case of IDF there is clear difference in the peaks of the signal. The demagnetized poles of the PM has lower magnetic field compared to the healthy one. Therefore, amplitude of the induced voltage signal at the demagnetized poles are lower compared to the healthy one as highlighted in Fig. 10(b).
Hence, a simple peak detection and/or calculating the differential voltage can give us a clear indication of the IDF.
The number of reduced peaks can also tell us that how many poles of PMs are demagnetized.
Similarly, in case of the static EF, the airgap at some portion is small and other portion is big. As a result, the PASC installed at lower airgap induce more voltage
(a)
(b)
Fig. 7. PASC. (a) Manufactred PASC, (b) Three PASCs installed at the pole shoes of the benchmark IPMSM.
0 90 180 270 360
-8 -6 -4 -2 0 2 4 6 8
500rpm 1000rpm 2000rpm 3000rpm 3500rpm
PSC Signal (V)
Electrical Angle (deg)
Fig. 8. Experimental result of the induced voltage at different speeds by PASC in the IPMSM's generating mode.
(a)
(b)
(c)
Fig. 9. Experimental result of the PASC. (a) Healthy, (b) IDF, (c)EF.
compared to the one installed at bigger airgap. This phenomenon can be seen in Fig. 10(c) where sensor 1 produced higher voltage while sensor 2 produced lower voltage. If we compare the magnitudes of the two sensor it will give us a significant difference which will be a clear sign of the EF. Hence, this kind of differential voltage can be used as EF fault indicator.
C. Vibration based Fault Signatures
Vibration signals contain the most important information and has been widely used for fault diagnosis in industry.
Vibration signals are obtained using accelerometer which has very high sensitivity and has the ability to capture the smallest change in the pattern of vibration. The vibration caused by different types of faults in a machine have different magnitude and different pattern. The electrical and magnet related faults may be classified using electrical signal. However, mechanical faults such as BF, which is the most frequently occurring fault, cannot be detect when it is at earliest stage. For instance, the electrical stress due to PWM inverter slowly damaging the surface of the bearing and normally this damage is uniformly applied to the entire surface. Therefore, it is very difficult to see any significant change in the stator current. However, such change in the surface of the bearing can easily be identified from the vibration signal. Fig. 11 shows the frequency spectrum and statical analysis of the time domain signal of the vibration obtained for healthy, BF and mix fault. It can be seen that there is a clear difference in the frequency and the RMS values, kurtosis, skewness, and standard deviation. It is worth mentioning that the severity of BF was very low in our case. The fault in the bearing was inserted by electrical
stress. In case of physical damage such as hole or fracture there will be huge change in vibration signal. However, the goal of this study is to analyze each fault with smallest severity.
I. DETECTION AND SEPARATION OF FAULTS A. Fault Diagnosis Algrithm
The proposed fault diagnosis uses the hybrid signature of MCSA, induced voltage of the PASC, and vibration signal for the classification of four types of faults. These The fault signature of these three signals complements each other that significantly reduce the risk of false alarm. The stator current, induced voltage of PASC, and vibration signals are obtained in the first step by experiments. The experimental setup is shown in Fig. 12 and the flowchart of the entire algorithm is presented in Fig. 13. FFT of current and vibration signals are calculated at 60% load and 3000rpm. In the meantime, the induced voltage of the PASC’s signals were also obtained and the peaks and differential voltages are calculated as well. The features extracted from all the signals are checked according to the rules given in Table II.
If any of the feature given in Table II exist that means that there is some issue in the motor. Thus, a general false alarm will be activated. Soon after the general fault detection, the fault indicator (FI) obtained are checked in detail and the classification of the fault is done based on the combination of the signature as per rules given in Table II.
• FI-1: FI-1 is the flag use for IDF in this study. If there is a significant 2nd and 4th order harmonic in the stator current and a reduction in the 5th order harmonic. This
0 1000 2000 3000 4000 5000
-80 -60 -40 -20 0 20 40 60 80
0 1000 2000 3000 4000 5000
-80 -60 -40 -20 0 20 40 60 80
0 1000 2000 3000 4000 5000
-80 -60 -40 -20 0 20 40 60 80
Vibration (mm/sec)
Frequency (Hz)
Healhty RMS value = 86.8 mm/s
Kurtosis = 1.997 Skewness = 0.224 std = 40.1
Vibration (mm/sec)
Frequency (Hz)
(a)
BF RMS value = 135.7 mm/s
Kurtosis = 2.23 Skewness = 0.0726 std = 60.5
Vibration (mm/sec)
Frequency (Hz)
(C) (b)
BF plus EF RMS value = 125.5 mm/s
Kurtosis = 2.34 Skewness = 0.143 std = 56.1
Fig. 10. Experimental results of the frequency specturm of vibration. (a) healthy, (b) BF, (c) mix fault (BF plus EF).
Table II. Fault indicator (FI) combinations.
Fautlt Type
MCSA PASC Signals Vibration
signal IDF
(FI-1)
2fs and 4fs
Reduction in 5fs
Mismatched Peaks and dsitorted shape of signal
FFT Kurtosis RMS Std Skewness Signifiant peaks in
differiental voltage of PASC signal.
ITSF (FI-2)
3fs and 9fs Reduction in the amplitude of one PASC sensor and increase in other.
FFT Kurtosis RMS EF
(FI-3)
X Different amplitudes of sensors with symetrical shapes
FFT Kurtosis RMS Std Skewness Large differiental voltage
between two sensors BF
(FI-4)
X Distorted shape of PASC signal.
FFT, Kurtosis, RMS, std, Mix Fault
(FI-5)
3fs (small) No clear / No fix sign FFT, Kurtosis, RMS, std
This clearly indicates partial IDF. Assume that the amplitudes of these harmonics are unclear or inconsistent. We can further examine the PASC's peaks and differential voltage. If some PASC peaks are lower than others, or if there are substantial peaks in the healthy and real-time differential voltage as shown in Fig. 12. This is yet another indication of the presence of IDF. This can be confirmed further by examining the frequency spectrum statical component of the vibration signal. The number of peaks with lower amplitude in the PASC signal indicates the number of PMs affected by IDF. This demonstrates that the suggested approach has the highest accuracy which significantly reduce the probability of false alarm.
• FI-2: FI-2 is the flag used for ITSF. If the stator current has substantial 3rd and 9th order harmonics. That is the sign of ITSF. It can be verified further by examining the PASC signals. If the magnitude of one PASC's signal decreases and there is an increase in vibration with particular harmonics, this indicates ITSF. The location of the ITSF can also be determined simply by observing which sensor has a lower amplitude than the others.
• FI-3: The next case of static EF. Under static eccentricity, there was no substantial and persistent harmonic observed in the stator current. If there is a considerable discrepancy in the amplitude of the two PASCs, this indicates that there is some kind of eccentricity. EF produces audible vibration along with additional harmonics. As a result, based on the PASC differential voltage (Fig. 12) and the vibration signal, it is easy to conclude that there is an eccentricity defect. The percentage reduction and increase in the PASC sensors can be used to quantify the severity of the EF. This fault signature combination is categorized as FI-3 here.
• FI-4: BF is the next case. If there is no severe damage and merely a rough surface or lubricant meltdown as a result of electrical stress in a bearing. There will be no clear signatures in the stator current. Furthermore, the
PASC showed no pattern of change other than a distorted form and a slight amplitude mismatch. As a result, BF bases its decision only on the signatures of the vibration signal. If there is no discernible change in the current signal or PASC, yet there is still huge vibration and acoustic noise. That is a clear indication of BF. This is classified as FI-4.
• FI-5: The final case is of a mix type of fault. The mix fault might be any of the above-mentioned defects combined. The combination of the BF and EF is investigated here as an example. A modest 3rd order harmonic in the stator current was measured in this case.
Along with the vibrations, there was a mismatch and deformed shape of the PASC. It is extremely difficult to pinpoint the specific pattern of the mix problem.
However, a mix kind of signature that is not one of the combinations described in Table II is a clear indication of a mix error. This combination is designated as FI-5.
B. Discussions
The proposed method incorporated the best features of each signal to broaden the diagnostics' scope in terms of fault type and redundant accuracy check. There is no doubt that the proposed method necessitates the use of vibration and PASC sensors, which may incur additional costs.
However, the vibration sensor is already a part of any diagnosis toolbox configuration in the industry, and the PASC is relatively inexpensive to consider from a cost standpoint. Furthermore, as compared to machine/deep learning-based categorization techniques, the suggested method is quite simple and does not require a specific processor or hard programming. Furthermore, the same fault signature can be used to determine the severity of multiple comparable faults and can estimate the severity as well.
Fig. 11. Experimental setup used in this study.
0 5 10 15 20 25 30 35
-6 -4 -2 0 2 4 6
Diff. voltage [V]
Time [ms]
Healthy, EF, IDF
Fig. 12. Differiental voltage obtained by PASC for Healthy, EF, and IDF.
Fig. 13. Flow chart of the proposed fault diagnosis algorithm.
II. CONCLUSION
A hybrid signature based fault diagnosis technique for PMSM is presented in this paper. The features of stator current, a planer-shaped search coil, and vibration signals are combined to diagnose ITSF, IDF, EF, BF, and mix fault. The combination of these signals improved the accuracy and reliability remarkably and also cover the most frequently occurring faults. This algorithm can also find the location of ITSF, number of poles demagnetized, and severity of EF.
The computational burden of this method is very low compared to those method based on machine/deep learning and can be easily implemented on a normal processor. In future work, more faults will be diagnose using this algorithm. In addition, the severity estimation and remaining useful life estimation will also be added.
REFERENCES
[1] S. Choi, E. Pazouki, J. Baek, and H. R. Bahrami, “Iterative condition monitoring and fault diagnosis scheme of electric motor for harsh industrial application,” IEEE Transactions on Industrial Electronics, vol. 62, no. 3, pp. 1760–1769, 2014.
[2] Z. Ullah, K.-T. Kim, J. Park, and J. Hur, “Comparative analysis of scalar and vector control drives of IPMSM under inter-turn fault condition considering nonlinearities,” in 2015 IEEE Energy Conversion Congress and Exposition (ECCE), 2015, pp. 366–372.
[3] Z. Ullah and J. Hur, “Irreversible Demagnetization Fault Prognosis in a Permanent Magnet type Machines,” in 2020 IEEE Energy Conversion Congress and Exposition (ECCE), 2020, pp. 742–747.
[4] Z. Ullah, S.-T. Lee, M. R. Siddiqi, and J. Hur, “Online diagnosis and severity estimation of partial and uniform irreversible demagnetization fault in interior permanent magnet synchronous motor,” in 2019 IEEE Energy Conversion Congress and Exposition (ECCE), 2019, pp. 1682–1686.
[5] H. Dehghan, F. Haghjoo, and S. M. A. Cruz, “A flux-based differential technique for turn-to-turn fault detection and defective region identification in line-connected and inverter-fed induction
motors,” IEEE Transactions on Energy Conversion, vol. 33, no. 4, pp.
1876–1885, 2018.
[6] Z. Ullah and J. Hur, “A Comprehensive Review of Winding Short Circuit Fault and Irreversible Demagnetization Fault Detection in PM Type Machines,” Energies (Basel), vol. 11, no. 12, p. 3309, 2018.
[7] M. Zafarani, E. Bostanci, Y. Qi, T. Goktas, and B. Akin, “Interturn short-circuit faults in permanent magnet synchronous machines: An extended review and comprehensive analysis,” IEEE Journal of Emerging and Selected Topics in Power Electronics, vol. 6, no. 4, pp.
2173–2191, 2018.
[8] B. M. Ebrahimi and J. Faiz, “Feature extraction for short-circuit fault detection in permanent-magnet synchronous motors using stator- current monitoring,” IEEE Transactions on Power Electronics, vol.
25, no. 10, pp. 2673–2682, 2010.
[9] B. M. Ebrahimi, J. Faiz, and M. J. Roshtkhari, “Static-, dynamic-, and mixed-eccentricity fault diagnoses in permanent-magnet synchronous motors,” IEEE Transactions on Industrial Electronics, vol. 56, no. 11, pp. 4727–4739, 2009.
[10] K.-T. Kim, S.-T. Lee, and J. Hur, “Diagnosis technique using a detection coil in BLDC motors with interturn faults,” IEEE Trans Magn, vol. 50, no. 2, pp. 885–888, 2014.
[11] J.-K. Park, C.-L. Jeong, S.-T. Lee, and J. Hur, “Early detection technique for stator winding inter-turn fault in BLDC motor using input impedance,” IEEE Transactions on Industry Applications, vol.
51, no. 1, pp. 240–247, 2015.
[12] H. Liang, Y. Chen, S. Liang, and C. Wang, “Fault detection of stator inter-turn short-circuit in PMSM on stator current and vibration signal,” Applied Sciences, vol. 8, no. 9, p. 1677, 2018.
[13] B. M. Ebrahimi and J. Faiz, “Demagnetization fault diagnosis in surface mounted permanent magnet synchronous motors,” IEEE Trans Magn, vol. 49, no. 3, pp. 1185–1192, 2012.
[14] Z. Ullah, S.-T. Lee, and J. Hur, “A novel fault diagnosis technique for IPMSM using voltage angle,” in 2018 IEEE Energy Conversion Congress and Exposition (ECCE), 2018, pp. 3236–3243.
[15] Z. Ullah, S.-T. Lee, and J. Hur, “A Torque Angle-Based Fault Detection and Identification Technique for IPMSM,” IEEE Transactions on Industry Applications, vol. 56, no. 1, pp. 170–182, 2019.
[16] J. Hong et al., “Detection and classification of rotor demagnetization and eccentricity faults for PM synchronous motors,” IEEE Transactions on Industry Applications, vol. 48, no. 3, pp. 923–932, 2012.
[17] R. Z. Haddad and E. G. Strangas, “Detection of static eccentricity and turn-to-turn short circuit faults in permanent magnet synchronous AC machines,” in 2015 IEEE 10th International Symposium on Diagnostics for Electrical Machines, Power Electronics and Drives (SDEMPED), 2015, pp. 277–283.
[18] Z. Ullah, B. A. Lodhi, and J. Hur, “Detection and identification of demagnetization and bearing faults in PMSM using transfer learning- based VGG,” Energies (Basel), vol. 13, no. 15, p. 3834, 2020.
[19] Z. Zhang, G. Luo, Z. Zhang, and X. Tao, “A hybrid diagnosis method for inverter open-circuit faults in PMSM drives,” CES Transactions on Electrical Machines and Systems, vol. 4, no. 3, pp. 180–189, 2020.
[20] I.-H. Kao, W.-J. Wang, Y.-H. Lai, and J.-W. Perng, “Analysis of permanent magnet synchronous motor fault diagnosis based on learning,” IEEE Transactions on Instrumentation and Measurement, vol. 68, no. 2, pp. 310–324, 2018.
[21] J.-H. Im and J. Hur, “Proposing New Planar-Type Search Coil for Permanent Magnet Synchronous Motor: Design and Application for Position Estimation,” IEEE Access, vol. 9, pp. 129078–129087, 2021.
[22] Z. Ullah and J. Hur, “Analysis of inter-turn-short fault in an FSCW IPM type brushless motor considering effect of control drive,” IEEE Transactions on Industry Applications, vol. 56, no. 2, pp. 1356–1367, 2019.
[23] J.-K. Park, T. R. Wellawatta, Z. Ullah, and J. Hur, “New equivalent circuit of the IPM-type BLDC motor for calculation of shaft voltage by considering electric and magnetic fields,” IEEE Transactions on Industry Applications, vol. 52, no. 5, pp. 3763–3771, 2016.
