Professional Documents
Culture Documents
Manuscript ID TII-20-0610
http://www.ieee-ies.org/
Page 1 of 8 Industrial Electronics Society
1
2
3
A Simple and Efficient Current Based Method
4
5 for Inter-turn Fault Diagnosis of Brush-less
6
7
8
Direct Current Motors
9
10 .
11
12
13 [11] to achieve the desired results. For fault magnetic field
Abstract—Fault diagnosis of brushless direct current (BLDC)
14 motors has recently gained more attention. Inter-turn faults in
variations can be used, which induces voltage in the search
15 coils as input. Although using the search coils lead to high
BLDC motors distort three-phase currents waveforms, leading to
16 an asymmetrical condition. This paper introduces a simple and accuracy, it is not cost-efficient in low-power motors.
17 efficient method to identify inter-turn faults based on one modal Moreover, the method cannot be considered as a general
18 current and four different simple indices. The modal current is technique for different types of BLDC motors. A parameter
19 derived by proper linear mixing of measured three-phase estimation-based technique has been introduced in [11]. In this
currents. Following the initial processing of the modal current, mathematical-based method, inverter voltage, dc current, and
20 three main indices including moving mean, variance, and signal rotor angular speed must be measured to diagnose the fault.
21 energy are obtained in parallel. Also, an auxiliary correlation- Inter-turn fault can be also diagnosed by input impedance
22 based index is suggested to enhance the method for [12]. Although the method has some advantages, high
23 discrimination of faulty conditions from non-faulty ones. The
computational burden and high demand of inputs are its main
24 fault detection is made by passing at least two main indices (out
drawbacks. The scheme needs rotation speed, input current,
25 of three indices), and also an auxiliary index from a predefined
threshold. The proposed technique is evaluated under different input voltage, resistance, and inductance to detect the fault,
26 and all these render it difficult and almost impractical to
loads, speeds, and fault severities in two different dataset: data
27 from a simulated motor in Maxwell finite element package, and a implement. Electromagnetic torque signal has been utilized
28 real 4-pole motor. Moreover, the method is compared with other for inter-turn fault diagnosis in different BLDC motors. These
29 methods from different aspects. The results confirm a high techniques identify the faults with the application of wavelet
30 accuracy and speed of the fault detection in the proposed transform [13], neuro-Fuzzy [14], and adaptive neuro-Fuzzy
31 approach. inference systems (ANFIS) [15]. Since the electromagnetic
Index Terms — Brushless direct current motor, inter-turn
32 torque signal is not easily accessible, these methods cannot be
fault, modal current.
33 easily implemented. Stator input current frequency pattern can
34 I. INTRODUCTION
also be generally utilized for inter-turn fault diagnosis [16-18].
35 Since any fault follows a definite frequency pattern, such
36
37
W ITH the development of industrial technologies,
brushless direct-current (BLDC) motors have attracted
interests as a large drive system in many industrial
pattern can be used for fault diagnosis. However, in BLDC
motors the harmonics frequency depends on the speed, so it is
38 difficult to identify the current spectrum for the fault pattern
fields. These motors manifest higher efficiency, higher power
39 during motor operation.
factor, and higher torque compared to other motors.
40 This paper proposes a novel technique to detect the inter-
Considering their wide applications, on-line condition
41 monitoring and fault diagnosis of BLDC motors are essential turn faults in BLDC motors using a multi-index current-based
42 for enhancement of reliability and continuous operation of the method. The main processed signal in the suggested method is
43 system [1-3]. a modal current which is based on linear mixing of measured
44 There are different types of faults that can be occured in three-phase currents. This signal shortens the computation
45 BLDC motors, and one of the major ones is the stator winding time and lessens the required memory. To identify the inter-
46 inter-turn short circuit (SC) which is caused by insulation turn faults, three different indices including moving mean,
47 defect. This fault is extremely harmful to BLDC motors variance, and signal energy are applied to the modal current in
48 because it can rapidly grow and result in motor failure [4-6]. parallel. Moreover, to distinguish the fault conditions from
49 That is why quick and accurate on-line diagnosis of this fault non-faulty ones, such as load change, an auxiliary correlation
50 is vital. In practice, there are some crucial features needed for based index is considered. To decide whether the fault occurs
51 on-line inter-turn fault diagnosis in BLDC motors, including: in BLDC motors, at least two indices (out of three) and the
52 1) Minimum the number of sensors for the proposed signal auxiliary index must exceed a predefined threshold.
53 sensing to lower the drive cost. Applications of these indices result in a fast and reliable fault
54 2) Utilizing a method with low computation burden to merge detection method. To evaluate the method in different
55 the drive system and the condition monitoring system. conditions, such as various loads, speeds, and fault severities,
56 3) Choosing a reliable and precise fault diagnosis method. two different datasets, including a finite element-based (FE)
57 Many techniques have been so far proposed for fault diagnosis simulation motor and a 4-pole BLDC motor, are considered.
58 of BLDC motors [7-10]. Search coil sensors have been used in Furthermore, the performance of the method is compared with
59
60 http://www.ieee-ies.org/
Industrial Electronics Society Page 2 of 8
1 some other similar methods from different aspects. the different coefficients, there is no direct subtracting or
2 To contributions of the proposed method include: summation of two signals. Consequently, it is guaranteed that
3 ● Application of modal current for inter-turn fault detection the transient behavior of the signals is unchanged and modal
4 in BLDC motors, and processing just one signal instead of signal preserves all the transient data in different conditions.
5 three current signals. Using the modal signal reduces the processing time and
6
● Simple implementation procedure as the algorithm only simplifies the development of the proposed algorithm. Fig. 1
7
requires simple operators. represents a typical three-phase current and its normalized
8
● Small computation burden and cost-efficient. modal signal for a typical inter-turn fault.
9
● Taking into account different challenging conditions in
10 C. 3rd Step: Calculating the Proposed Indices
evaluating the method.
11 To make suitable criteria for the inter-turn fault diagnosis in
● Having relatively high accuracy and speed.
12 BLDC motors, three different simple mathematical-based
13 indices are proposed. It should be mentioned that all of these
II. PROPOSED METHOD
14 indices are computed in parallel. In the following subsections,
15 The proposed method employs the measured three-phase
each index is explained more in details.
16 current signal as the main signal to identify the faults. At the
17 beginning, the measured currents are normalized and then the C.1. Moving Average-based Index
18 modal current is built up based on these normalized signals. The first proposed index is the moving average-based one.
19 Having the modal current, three different moving indices The moving averaging for a sample discrete signal x(t), for a
20 including mean-based index (MBI), variance-based index window with T0 length, is as follows [22, 23]:
21 (VBI) and energy-based index (EBI) are calculated in parallel
1 k
22 to recognize the inter-turn fault condition. Application of these MA k
T0
x k (t ) (3)
23 three indices enables the method to investigate the signal from n k T 0
24 three different aspects, and increase the reliability and where MAk is the mean of signal x(t) in kth window. In the
25 quickness of the fault diagnosis process. Also, these indices proposed method, moving average is applied to the modal
26 are easily computed by simple mathematical operators. Hence, current signal S, with the number of samples in one cycle
27 the method can be easily implemented online. To differentiate designated as Ts, and the time step of one sample designated as
28 the non-faulty conditions, such as load change, from faulty t. So, the proposed moving mean-based index (MBI) is derived
29 ones, an auxiliary index is also computed. The detailed as follows:
30 description of the method are as follows: 1 k
31 A. 1st Step: Signal Normalizing
MBI k S k (t )
T S n k TS
(4)
32 The value of the proposed MBI is almost unchanged in the
Amplitude of the current signals differs at different loads.
33 motors but it suddenly changes following the inter-turn fault.
As a result, to simplify the signal processing, the signals must
34 Therefore, the MBI is compared with a threshold to decide
be expressed in per unit (pu) to limit the range of amplitude of
35 whether this index flags the fault occurrence or not. The value
all signals (i). This is done by dividing all samples to the peak
36 of the threshold for this index is considered ±0.22, using a
current (im) as follows:
37 straight-forward algorithm named “Otsu thresholding method”
38 i
i pu (1) [24].
39 im
40 C.2. Moving Variance-based Index
This action places all signals in the range of -1 and 1. So, the
41 Moving variance-based index is the second proposed index
currents of motor are easily considered under different
42 to distinguish the inter-turn fault in BLDC motors. Knowing
operating conditions.
43 the formula for moving variance, the proposed variance-based
44 B. 2nd Step: Transforming Normalized Current Signals to index (VBI) is derived as follows:
45 Modal Signal
46 Since inter-turn faults may occur in different phases, all
47 three phases of the motor must be taken into consideration. By
48 transforming three-phase currents to a modal signal, the
49 computation time is shortened and less memory is required
50 [19, 20]. Consequently, instead of three-phase signals, the
51 modal signal is processed. The modal signal is a linear
52 combination of the three-phase currents as follows:
53 S S a S b S c (2)
54
where S is the current modal signal, Sa, Sb, and Sc are the
55
normalized signals of three-phase currents respectively. Also
56
α, β, and γ are the modal signal coefficients considered as 1, 2,
57 Fig.1. Three-phase currents and the modal current for an inter-turn fault.
and -3, respectively [21]. Since the phases are multiplied by
58
59
60 http://www.ieee-ies.org/
Page 3 of 8 Industrial Electronics Society
54 t t (7)
55 ( x i x ) 2 ( y i y ) 2
t T 0 t T 0
56 where cxy denotes the correlation between variables x and y, T0 Fig. 2. Flowchart of proposed method.
57 is the length of data window, t is the sample number, while μx
58
59
60 http://www.ieee-ies.org/
Industrial Electronics Society Page 4 of 8
1
2
3 DC
4 Voltage
Source
5
6 Motor’s
speed
7
Desired
8 Switches’
gates signal
PWM PI
controller speed
9 (a)
Fig. 6. Schematic of designed closed-loop driver.
10
TABLE II.
11
SPECIFICATIONS OF BLDC MOTOR
12
13 Parameter Value
14 Rated Power (W) 370
15
Voltage (V) 48
16
Current (A) 10.5
17
18 (b) Number of poles 4
19 Rated speed (rpm) 3000
20 Rated torque (Nm) 1.2
21 Maximum torque (Nm) 3.6
22 Length (mm) 115
23
Weight (kg) 2.5
24
25
TABLE III.
26 SPECIFICATIONS OF THE DRIVER AND THE TEST BENCH DEVICES
(c)
27
28 Parameter Value
29 DC voltage source (V) 48
30 Current probe GWINSTEK GCP-100
31 Microcontroller DSPIC33FJ128MC708
Switch IRFP150N MOSFET
32
Frequency (kHz) 5
33 Switch driver IR2130
34
35 (d) Considering the permissible current of the rheostats, the
36 tests were carried out with maximum current 2 A.
37 Here, data of more than 30 different test cases are gathered.
38 It should be mentioned that current waveforms were gathered
39 using a data logger with the sampling frequency of 5 kHz.
40
B. Experimental results
41
42 In this subsection, the proposed method is assessed by some
43 gathered experimental data from different experiments. Fig. 7
(e)
44 Fig. 4. Inputs and outputs of proposed method for a fault with 6 SC turns in
shows the typical fault current waveforms for different
45 simulated BLDC motor; (a) three phase currents (b) MBI index (c) VBI index
severities. As shown in this figure, changes in the amplitude
46 (d) EBI index and (e) CBI index. and the waveform of the faults with higher severities are more
47 significant compared with the healthy operation mode of the
48 motor. In other words, the faults with lower severities are
49 much harder to identify. So, this type of inter-turn faults
50 (which has more probability of occurrence than high-severity
51 faults) is considered here to be studies in more details. Further
52 on, a typical low-severity inter-turn fault is investigated.
53 Fig. 8 shows the current waveforms and the proposed
54 indices for practical fault data. It should be mentioned that this
55 specifically studied fault was occurred at 10% of the nominal
56 load and 40% of the nominal speed. The depicted MBI in Fig.
57 8(b) shows that this index has failed detect the fault. This issue
58 Fig.5. Practical test bench for experimental verification.
59
60 http://www.ieee-ies.org/
Industrial Electronics Society Page 6 of 8
1 TABLE IV.
2 COMPARATIVE ASSESSMENT OF PROPOSED METHOD
3
4 Parameter
Method Method Method Proposed
5 A [23] B [12] C [17] method
6 Accuracy (%) 67.35 72.86 87.25 97.56
7
8 Online capability Yes Yes No Yes
9 Detectability of
10 lowest severity 4 4 5 3
(a)
11 faults (SC turns%)
12 Required average
13 time to operate 31 34 26 16
(ms)
14
Discrimination
15 capability of non- No No No Yes
16 faults
17 Requirement of
18 additional sensors Yes No No No
19 (b)’ or hardware
20 Computational
Low High High Low
21 Fig. 9. Measured currents and its modal current for a typical load change. burden
22 of the BLDC motors [17]. In the proposed methodology,
23 compared to other harmonics, generation of the 3rd harmonic
24 is greater. So, 3rd harmonic in the negative frequency (–3f) is
25 considered as the fault detector criterion.
26 In the following, the proposed technique is compared with
27 the aforementioned methods based on seven criteria including
28 accuracy, online fault detection capability, detectability of
29 lowest severity faults (SC turns%), required average time for
(a)
30 fault detection, discrimination capability of faults from load
31 change, requirement of additional sensors or hardware, and
32 computational burden. The total number of these case studies
33 is more than 90.
34
In Table IV tabulates the results. It is obvious that the
35
proposed method can successfully detect the faults, while the
36
operation in other methods is disfunctional in some cases.
37
Moreover, the average time delay of the proposed scheme is
38
(b) about 16 ms which is longer than the other algorithms. The
39
proposed algorithm only needs three current sensors which are
40
more economical compared to three other methods, it is
41
capable to merge the drive and fault diagnosis system. Also,
42
the computational burden of the proposed method is
43
considerably low compared with the other methods. The other
44
criteria indicate that the proposed algorithm has high degrees
45
of superiority over other methods.
46
47 (c) IV. CONCLUSION
48
49 This paper proposed a simple and accurate technique to
50 detect the inter-turn fault of BLDC motors. The method
51 utilizes three main indices for identifying the change occurred
52 in the motor, and an auxiliary index is suggested to
53 discriminate the faulty conditions from non-faulty ones. The
54 proposed algorithm is evaluated in FE-based simulated BLDC
55 motor and experimental test bench. The results confirmed that
56 (d) the proposed method can successfully handle critical
57 Fig. 10. Indices of proposed method for a load change in test bench; (a) MBI conditions, such as load change. Also, the achieved results
index (b) VBI index (c) EBI index and (d) CBI index.
58 denoted that the algorithm can be easily implemented online,
59
60 http://www.ieee-ies.org/
Industrial Electronics Society Page 8 of 8
1 which only requires around 16 ms for identification and is a [20] T. Ghanbari, "Autocorrelation function-based technique for stator turn-
2 low calculation demand scheme with high reliability. fault detection of induction motor," IET Science, Measurement &
3 Technology, vol. 10, pp. 100-110, 2016.
[21] N. Perera, A. Rajapakse, and T. Buchholzer, "Isolation of faults in
4 REFERENCES distribution networks with distributed generators," IEEE Trans. on
5 Power Delivery, vol. 23, pp. 2347-2355, 2008.
[1] K.-T. Kim, Y.-S. Lee, and J. Hur, "Transient analysis of irreversible
6 demagnetization of permanent-magnet brushless DC motor with
[22] M. A. Jarrahi, H. Samet, and T. Ghanbari, "Fast Current-Only Based
Fault Detection Method in Transmission Line," IEEE Systems Journal,
7 interturn fault under the operating state," IEEE Trans. on Industry pp. 1-12, 2018.
8 Applications, vol. 50, pp. 3357-3364, 2014. [23] K.-T. Kim, S.-T. Lee, and J. Hur, "Diagnosis technique using a detection
[2] Munikoti, Sai, Laya Das, Balasubramaniam Natarajan, and Babji
9 Srinivasan. "Data-Driven Approaches for Diagnosis of Incipient Faults
coil in BLDC motors with interturn faults," IEEE Trans. on Magnetics,
10 in DC Motors." IEEE Transa. on Industrial Informatics, vol. 15, no. 9,
vol. 50, pp. 885-888, 2014.
[24] N. Otsu, "A threshold selection method from gray-level histograms,"
11 pp. 5299-5308, 2019. IEEE Trans. on Systems, Man, and Cybernetics, vol. 9, pp. 62-66, 1979.
12 [3] Yu, Ming, Chenyu Xiao, Wuhua Jiang, Shuanglong Yang, and Hai [25] E. Mazaheri-Tehrani, J. Faiz, M. Zafarani, and B. Akin, "A Fast Phase
Wang. "Fault diagnosis for electromechanical system via extended
13 analytical redundancy relations." IEEE Trans. on Industrial
Variable abc Model of Brushless PM Motors Under Demagnetization
Faults," IEEE Trans. on Industrial Electronics, vol. 66, pp. 5070-5080,
14 Informatics, vol. 14, no. 12, pp. 5233-5244, 2018. 2019.
15 [4] M. Awadallah and M. Morcos, "Diagnosis of stator short circuits in
brushless dc motors by monitoring phase voltages," IEEE Trans. on
16 Energy Conversion, vol. 20, pp. 246-247, 2005.
17 [5] Aqil, Muhammad, and Jin Hur. "A Direct Redundancy Approach to
18 Fault-Tolerant Control of BLDC Motor With a Damaged Hall-Effect
Sensor." IEEE Trans. on Power Electronics, vol. 35, no. 2, pp. 1732-
19 1741, 2019.
20 [6] Hosseini, Saba M., Faezeh Hosseini, and Mostafa Abedi. "Stator fault
21 diagnosis of a BLDC motor based on discrete wavelet analysis using
ADAMS simulation." SN Applied Sciences, vol. 1, no. 1, pp. 1406-1418,
22 2019.
23 [7] J. Faiz and A. Jafari, "Interturn fault diagnosis in brushless direct current
24 motors—A review," 2018 IEEE International Conference on Industrial
25 Technology (ICIT), 2018, pp. 437-444.
[8] Zandi, Omid, and Javad Poshtan. "Fault diagnosis of brushless DC
26 motors using built-in hall sensors." IEEE Sensors Journal, vol. 19, no.
27 18, pp. 8183-8190, 2019.
28 [9] Medeiros, Ramon LV, Abel C. Lima Filho, Jorge Gabriel GS Ramos,
Tiago P. Nascimento, and Alisson Vasconcelos Brito. "A novel
29 approach for speed and failure detection in brushless DC motors based
30 on chaos." IEEE Tran. on Industrial Electronics, vol. 66, no. 11, pp.
31 8751-8759, 2018.
[10] Ullah, Zia, and Jin Hur. "Analysis of Inter-Turn-Short Fault in an FSCW
32 IPM type Brushless Motor Considering Effect of Control Drive." IEEE
33 Trans. on Industry Applications, DOI: 10.1109/TIA.2019. 2961878,
34 2019.
35 [11] K.-T. Kim and J. Hur, "Accurate and simple diagnosis algorithm for
inter-turn fault in the BLDC motor," 2013 IEEE Energy Conversion
36 Congress and Exposition, 2013, pp. 3534-3540.
37 [12] J.-K. Park, C.-L. Jeong, S.-T. Lee, and J. Hur, "Early detection technique
38 for stator winding inter-turn fault in BLDC motor using input
impedance," IEEE Trans. on Industry Applications, vol. 51, pp. 240-247,
39 2014.
40 [13] M. A. Awadallah, M. M. Morcos, S. Gopalakrishnan, and T. W. Nehl,
41 "Detection of stator short circuits in VSI-fed brushless DC motors using
wavelet transform," IEEE Trans. on Energy Conversion, vol. 21, pp. 1-
42 8, 2006.
43 [14] M. A. Awadallah, M. M. Morcos, S. Gopalakrishnan, and T. W. Nehl,
44 "A neuro-fuzzy approach to automatic diagnosis and location of stator
inter-turn faults in CSI-fed PM brushless DC motors," IEEE Trans. on
45 Energy Conversion, vol. 20, pp. 253-259, 2005.
46 [15] M. Awadallah and M. Morcos, "ANFIS-based diagnosis and location of
47 stator interturn faults in PM brushless DC motors," IEEE Trans. on
48 Energy Conversion, vol. 19, pp. 795-796, 2004.
[16] J.-K. Park, I.-M. Seo, and J. Hur, "Fault type detection using frequency
49 pattern of stator current in IPM-type BLDC motor under stator inter-
50 turn, dynamic eccentricity, and coupled faults," 2013 IEEE Energy
51 Conversion Congress and Exposition, 2013, pp. 2516-2521.
[17] S.-T. Lee and J. Hur, "Detection technique for stator inter-turn faults in
52 BLDC motors based on third-harmonic components of line currents,"
53 IEEE Trans. on Industry Applications, vol. 53, pp. 143-150, 2017.
54 [18] J.-K. Park and J. Hur, "Detection of inter-turn and dynamic eccentricity
faults using stator current frequency pattern in IPM-type BLDC motors,"
55 IEEE Trans. on Industrial Electronics, vol. 63, pp. 1771-1780, 2015.
56 [19] M. A. Jarrahi and H. Samet, "Modal Current and Cumulative Sum Based
57 Fault Detection Approach in Transmission Lines," International Journal
58 of Emerging Electric Power Systems, vol. 19, 2018.
59
60 http://www.ieee-ies.org/