2015 European Control Conference (ECC)

July 15-17, 2015. Linz, Austria

Demagnetization Fault diagnosis of PMSM based on Analytic

Inductance Calculation*
Seokbae Moon, Jewon Lee, and Sang Woo Kim, Member, IEEE

Abstract— Demagnetization fault of Permanent magnet

II. ELECTRICAL MODEL OF PMSM AND FLUX ESTIMATION
synchronous machine (PMSM) degrades motor’s performance.
As the strength of permanent magnet (PM) decreases, A. PMSM Model
inductance of the PMSM changes since flux density affects
electrical steel’s nonlinear characteristics. It makes difficult to The electrical model of PMSM in the synchronous
estimate exact flux linkages of the PMSM to diagnose reference frame is given by
demagnetization fault. In this paper, we suggested analytic 𝑣𝑑 𝑟𝑠 + 𝑝𝐿𝑑 −𝜔𝑒 𝐿𝑞 𝑖𝑑 0
inductance prediction considering magnetic saturation effect [𝑣 ] = [ ][ ] + [ ] (1)
𝑞 𝜔𝑒 𝐿𝑑 𝑟𝑠 + 𝑝𝐿𝑞 𝑖𝑞 𝜔𝑒 𝜓𝑃𝑀
and varying flux linkages of PMSM, which shows better results
in estimation of flux linkages. Finite Element Analysis (FEA) where 𝜔𝑒 is the electrical angular velocity, 𝑖𝑑 and 𝑖𝑞 are d-
results demonstrate that the proposed method works well. and q-axes armature currents, 𝑣𝑑 and 𝑣𝑞 are d- and q-axes
I. INTRODUCTION terminal voltages, 𝑟𝑠 is the armature resistance, 𝐿𝑑 and 𝐿𝑞 are
the d- and q-axes inductances, 𝜓𝑃𝑀 is the PM’s flux linkages,
The permanent magnet synchronous machine (PMSM) is 𝑑
widely used in many applications due to its high performance. and 𝑝 = .
𝑑𝑡
As there is no copper loss of secondary winding, it has higher Since the equation including 𝑣𝑑 has no relationship with
efficiency than induction motor [1]. Many advantages that flux linkages of PMSM, only the other equation which
include high power density, precise torque control, high torque concerns 𝑣𝑞 is used for flux estimation. Rearranged and
to volume ratio and high performance makes it attractive discretized equation of 𝜓𝑃𝑀 is given by
candidate for many industrial areas.
As a consequence, many researches have been done to 𝑤𝑒 (𝑘)𝜓𝑃𝑀 = 𝑣𝑞 (𝑘) − 𝑤𝑒 𝐿𝑞 𝑖𝑑 (𝑘) − 𝑟𝑠 𝑖𝑞 (𝑘)
𝑖𝑞 (𝑘)−𝑖𝑞 (𝑘−1)
maintain such high performance of the PMSM and even make −𝐿𝑞 + 𝑛(𝑘) (2)
𝑇𝑠
better. And fault diagnosis technique also has been studied
since faults can degrade the machine’s operation such as lower where 𝑇𝑠 is sampling time and n(k) is the model uncertainty at
torque, poor efficiency and so on. Early and correct diagnosis sequence k due to harmonics of current, electrical noise and so
of a fault in electrical drive ensures high reliability and safety on.
of the system. There are several kinds of faults can occur in
PMSM system: bearing faults, electrically short/open circuit, B. Flux Estimation via Least Square Method
eccentricity and demagnetization problem [2]-[4]. The Least Square method is the most popular parameter
Demagnetization fault means that strength of PM identification method. It gives optimal solution for moving
decreases. The fault is usually caused by physical damage, average model system with additive White Gaussian noise.
high temperature stress, inverse magnetic field and aging. If 𝐿𝑞 , 𝐿𝑑 , and 𝑟𝑠 are known exactly, then 𝜓𝑃𝑀 can be
There has been effort towards to detect demagnetization fault estimated through Least Square method and result is given by
involving motor current signature analysis (MCSA), following steps.
time-frequency analysis methods such as Fast Fourier 𝑖𝑞 (1)−𝑖𝑞 (0)
𝑣𝑞 (1) − 𝑤𝑒 𝐿𝑞 𝑖𝑑 (1) − 𝑟𝑠 𝑖𝑞 (1) − 𝐿𝑞 + 𝑛(1)
Transform (FFT) [5], continuous Wavelet Transform (CWT) 𝑇𝑠
[6] and Hilbert-Huang transform [4]. However, in many 𝑖𝑞 (2)−𝑖𝑞 (1)
𝑣𝑞 (2) − 𝑤𝑒 𝐿𝑞 𝑖𝑑 (2) − 𝑟𝑠 𝑖𝑞 (2) − 𝐿𝑞 + 𝑛(2)
practical cases, analysis of the signal features is hard and it 𝑇𝑠

needs heavy calculation load. ⋮

𝑖𝑞 (𝑚)−𝑖𝑞 (𝑚−1)
In this paper, PMSM is simulated with FEA at different 𝑣𝑞 (𝑚) − 𝑤𝑒 𝐿𝑞 𝑖𝑑 (𝑚) − 𝑟𝑠 𝑖𝑞 (𝑚) − 𝐿𝑞 + 𝑛(𝑚)
𝑇𝑠
demagnetization levels with varying load. A novel approach [ ]
for estimating flux linkages based on inductance calculation 𝑤𝑒 (1)
with structure analysis is suggested. The method allows us 𝑤𝑒 (2)
estimating flux linkages of PMSM more precisely. Simulation = ⋮ 𝜓𝑃𝑀 (3)
results demonstrate that proposed method can diagnose the 𝑤𝑒 (3)
fault of demagnetization well. [ ]
Y = Ω𝜓𝑃𝑀 (4)
*Research is supported by MSIP
Seokbae Moon and Jewon Lee Sang Woo Kim is with Department of 𝜓̂𝑃𝑀 = (Ω Ω) Ω 𝑌
𝑇 −1 𝑇
(5)
Electrical Engineering, POSTECH, Pohang, Gyeongbuk, Korea (e-mail:
msbworld, [email protected]). But it is hard to know exact value of 𝐿𝑞 , 𝐿𝑑 , and 𝑟𝑠 since
Sang Woo Kim is with Department of Electrical Engineering and creative magnetic saturation effect and varying temperature give
IT Engineering, POSTECH, Pohang, Gyeongbuk, Korea (e-mail:
[email protected]).

978-3-9524269-3-7 ©2015 EUCA 3186

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on October 14,2020 at 16:03:48 UTC from IEEE Xplore. Restrictions apply.
variety to inductance and resistance of PMSM. Incorrect
parameters may result poor demagnetization diagnosis.
III. ANALYTIC CALCULATIONS OF IPMSM INDUCTANCE
A. Relationship between d-q and Phase Inductances
It is known that flux linkages of the PMSM in a
synchronous d-q frame is given by
𝜆𝑑𝑞 = 𝐿𝑠 𝑖𝑑𝑞 + 𝜓𝑃𝑀 (6)
3 (a)
where 𝐿𝑠 = 𝐿𝑚𝑠 + 𝐿𝑙𝑠 . Denote by 𝐿𝑚𝑠 and 𝐿𝑙𝑠 stator mutual
2
inductance and stator leakage inductance. For Interior
permanent magnet synchronous machine (IPMSM), equation
(6) changes as follows
3
𝜆𝑑𝑞 = 𝐿𝑠 𝑖𝑑𝑞 − 𝐿𝛿 (𝑖𝑑𝑞 )∗ + 𝜓𝑃𝑀 (7)
2

As IPMSM does not have PM flux along q-axis, it has

reluctance component, described as 𝐿𝛿 in above equation.
D-axis and q-axis inductance of IPMSM are represented by
3 3
𝐿𝑑 = 𝐿𝑠 − 𝐿𝛿 and 𝐿𝑞 = 𝐿𝑠 + 𝐿𝛿 . Nominal PMSM flux (b)
2 2
dynamics is described as following equations. Figure1. (a) a-phase coil aligned with d-axis of IPMSM
(b) a-phase coil aligned with q-axis of IPMSM
𝜆𝑑 = 𝐿𝑑 𝑖𝑑 + 𝜓𝑃𝑀
(8)
𝜆𝑞 = 𝐿𝑞 𝑖𝑞
Extending this characteristic to phase inductance, we can
describe a-phase inductance as
𝐿𝑎 (𝜃) = 𝐿𝑚𝑠 − 𝐿𝛿 𝑐𝑜𝑠2𝜃 (9)
Usually 𝐿𝑙𝑠 is much less than 𝐿𝑚𝑠 in IPMSM, we may use
approximations in relationship between 𝐿𝑎 with 𝐿𝑑 and 𝐿𝑞 as
follows
3 3 3 3
𝐿𝑑 = 𝐿𝑠 − 𝐿𝛿 = (𝐿𝑚𝑠 − 𝐿𝛿 ) + 𝐿𝑙𝑠 ≈ (𝐿𝑚𝑠 − 𝐿𝛿 ) = 𝐿𝑎,𝑑
2 2 2 2
3 3 3 3
𝐿𝑞 = 𝐿𝑠 + 𝐿𝛿 = (𝐿𝑚𝑠 + 𝐿𝛿 ) + 𝐿𝑙𝑠 ≈ (𝐿𝑚𝑠 + 𝐿𝛿 ) = 𝐿𝑎,𝑞
2 2 2 2
(10)
where 𝐿𝑎,𝑑 and 𝐿𝑎,𝑞 stand for a-phase inductance with d-axis
aligned (Fig 1. a) and a-phase inductance with q-axis aligned
(Fig. 1. b). We can get 𝐿𝑑 and 𝐿𝑞 by calculating a-phase
Figure 2.Flux density distribution of IPMSM, zero current, given by FEA
inductance at specific moment of motor operation. simulation
B. Calculation of Phase Inductance
 Stator teeth, shoe and rotor’s bridge have varying
There have been many researches to estimate and calculate reluctance due to the varying current and strength of
inductance of the PMSM since exact parameters are necessary PM. The nonlinear characteristic of iron is described
to control the motor maintaining high performance [7], [8]. by electrical steel’s B-H curve given by FEA
Before analyzing magnetic circuit of PMSM, we simplified simulation tool, Maxwell.
motor structure (Fig. 1) such as teeth, shoe of stator and rotor’s
bridge. Further, some assumptions listed here are made to Conventional analytic approach calculates with stator teeth,
simplify the analysis. air-gap and magnet [7], [8]. In case of IPMSM, flux density of
bridge is especially high (Fig. 2) and results significant
 For the calculation of air-gap inductance, the field reluctance, considering rotor bridge’s inductance is necessary
pattern of flux is assumed to be orthogonal to the to figure out exact inductance.
intersection surface of the air-gap and steel. Starting from the magnetic equivalent circuit in the d
 The B-H curve of the PM is linear with the constants direction of the machine, considering two a-phase coils since
of the relative permeability and remanence. there exists closed magnetic loop between two PMs of
 The reluctance of the rotor and the stator back iron are IPMSM (Fig. 4). Denote each component of inductance by 𝐵𝑛 .
constant. Major contribution is given by air-gap (𝐿0 ), PM (𝐿1 ) and
rotor’s bridge (𝐿2 ) and minor is given by stator’s teeth (𝐿3 ),
stator back iron (𝐿4 ) and rotor inner iron (𝐿5 ) (Fig. 3).

3187

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on October 14,2020 at 16:03:48 UTC from IEEE Xplore. Restrictions apply.
 stator teeth due to current. Fluxes of PM 𝜙𝑥,𝑃𝑀 can be
calculated or could be acquired from the value of FEA result in
normal motor. We took a later method. It is easy to calculate
𝜙𝑡𝑒𝑒𝑡ℎ,𝑐𝑢𝑟 with the following equation [9].
𝜙𝑡𝑒𝑒𝑡ℎ,𝑐𝑢𝑟 = 𝜙𝑡𝑒𝑒𝑡ℎ,𝑖𝑎 + 𝜙𝑡𝑒𝑒𝑡ℎ,𝑖𝑏 + 𝜙𝑡𝑒𝑒𝑡ℎ,𝑖𝑐 (
𝐿𝑡𝑒𝑒𝑡ℎ
Figure 3.Equivalent magnetic circuit of a-phase (d-axis aligned) = (𝑖𝑎 + 𝑖𝑏 + 𝑖𝑐 ) (19)
2𝑁

Figure 5.Equivalent magnetic circuit of a-phase (q-axis aligned)

As you can see in Fig. 4, the motor has structural

characteristic that one coil contains five teeth and shoes. Each
of them get different affect by other coils, so final flux
linkages made by current is different from each other and it
makes different reluctance to each teeth and shoes. This
phenomenon results difference to equations (13), (14), which
are teeth and shoes’ inductance according to current
conditions. This also happens in rotor’s bridge area and gives
variations to equation (15), finally makes varying inductance
of IPMSM with different current conditions.
We can get the final value of phase inductance from
overall magnetic circuit. In view of flux vector, reluctances are
Figure 4.Field orientation of IPMSM, zero current, given by FEA connected in series way, so we can figure out inductance by
simulation parallelizing every inductance value as following equation.
We can calculate phase inductance from each component’s 𝐿𝑎,𝑑 = 6(𝐿0 //𝐿0 //𝐿1 ⋯//𝐿4 //𝐿5 ) (20)
inductance, which are derived by reluctance of structural
characteristics. We adopted 2D equivalent motor model. We As you can see in Fig. 3, 𝐿4 and 𝐿5 appears only once
simplified stator’s teeth and shoes as one component, 𝐿3 . where every other inductance component appears twice, so it
needs to be considered in above equation. As PMSM model
𝑁2 𝑔 has 6 series phase windings (Fig. 4), we multiplied 6 to
𝐿0 = , 𝑅0 = (11)
𝑅0 𝜇𝑎𝑖𝑟 (5𝑤𝑠 +4𝑏0 )d parallel form of given inductance. We can finally get
𝑁2 ℎ𝑃𝑀 inductance of d-axis with (10).
𝐿1 = , 𝑅1 = (12)
𝑅1 𝜇𝑃𝑀 (𝑤𝑃𝑀 +𝑤𝑎𝑖𝑟 )d In case of q-axis inductance, equivalent magnetic circuit is
𝑁2 ℎ𝑏𝑟𝑖𝑑𝑔𝑒
𝐿2 = , 𝑅2 = (13) simplified as Fig. 5. Instead of magnet’s inductance placement,
𝑅2 𝜇𝑠 (𝐵𝑏𝑟𝑖𝑑𝑔𝑒 )𝑤𝑏𝑟𝑖𝑑𝑔𝑒 d
there exists narrow path of steel in the path of q-axis. This path
𝑁2 ℎ𝑡
𝐿3 = , 𝑅3 = (14) is usually called ‘rib’ and its inductance is
𝑅3 𝜇𝑠 (𝐵𝑡𝑒𝑒𝑡ℎ)5𝑤𝑡 d
𝑁2 ℎ𝑠𝑡𝑎𝑡𝑜𝑟 𝑁2 ℎ𝑟𝑖𝑏
𝐿4 = , 𝑅4 = (15) 𝐿6 = , 𝑅6 = (21)
𝑅5 𝜇𝑠 𝑤𝑠𝑡𝑎𝑡𝑜𝑟 d 𝑅6 𝜇𝑠 𝑤𝑟𝑖𝑏 d
𝑁2 ℎ𝑟𝑜𝑡𝑜𝑟
𝐿5 = , 𝑅5 = (16) Similar approach to q-axis equivalent magnetic circuit (Fig.
𝑅5 𝜇𝑠 𝑤𝑟𝑜𝑡𝑜𝑟 d
5), we can get the phase inductance with q-axis aligned.
where N is the number of turns in stator’s coil, g and 𝑤𝑠 , ℎ𝑡 ,
𝑏0 are described in Fig.1, d is depth of motor, 𝑤𝑥 is width of x 𝐿𝑎,𝑞 = 6(𝐿0 //𝐿0 //𝐿3.𝑞 ⋯//𝐿4 //𝐿5 ) (22)
(x could be stator, PM, bridge, teeth and so on), ℎ𝑥 is height of Proposed approach made us to follow the change in
x, 𝑤𝑎𝑖𝑟 is width of air next to PM inside of rotor, and 𝜇𝑠 (𝐵𝑥 ) is inductance well, which will be shown in section Ⅴ.
absolute permeability of steel when flux density is 𝐵𝑥 . Flux
densities of each part is determined by strength of PM and IV. DEMAGNETIZATION FAULT DIAGNOSIS SCHEME
current. These flux densities are defined as follows
There have been many researches for demagnetization
𝜙𝑡𝑒𝑒𝑡ℎ,𝑃𝑀 +𝜙𝑡𝑒𝑒𝑡ℎ,𝑐𝑢𝑟
𝐵𝑡𝑒𝑒𝑡ℎ = 𝐵𝑡𝑒𝑒𝑡ℎ,𝑃𝑀 + 𝐵𝑡𝑒𝑒𝑡ℎ,𝑐𝑢𝑟 = (17) fault diagnosis and most of them use MCSA utilizing FFT,
𝑤𝑡 𝑑
short time Fourier transform (STFT), CWT and so on. These
𝜙𝑏𝑟𝑖𝑑𝑔𝑒,𝑃𝑀 +𝜙𝑏𝑟𝑖𝑑𝑔𝑒,𝑐𝑢𝑟 techniques diagnose that the motor has fault or not. However,
𝐵𝑏𝑟𝑖𝑑𝑔𝑒 = (18)
𝑤𝑏𝑟𝑖𝑑𝑔𝑒 𝑑 they couldn’t give the level of fault. There were other
approaches to give the degree of fault, but even they
where 𝜙𝑡𝑒𝑒𝑡ℎ,𝑃𝑀 and 𝜙𝑡𝑒𝑒𝑡ℎ,𝑐𝑢𝑟 are respectively, the fluxes of considered the variation of inductance, couldn’t figure out
PM flowing through stator teeth and slot linkages fluxes of result precisely [10].

3188

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on October 14,2020 at 16:03:48 UTC from IEEE Xplore. Restrictions apply.
 Fig. 6 shows the proposed fault diagnosis’ scheme. With TABLE I. PARAMETERS OF IPMSM IN SIMULATION

initial normal data of motor, we can estimate the degree of Parameters

demagnetization if the motor has fault. In step 1, we estimate
Pole pairs 3
𝜓𝑃𝑀 by Least Square method with normal value of 𝐿𝑑 and 𝐿𝑞 .
Rated current 11 A
In step 2, update 𝐿2 and 𝐿3 with estimated 𝜓̂𝑃𝑀 . Using normal
Rated torque 4.8 Nm
inductance values though, estimated 𝜓̂𝑃𝑀 shows change if
there were demagnetization fault [10]. In step 3, do d-axis inductance 2.019 [mH]
estimations with updated inductance 𝐿𝑑 ′ and 𝐿𝑞 ′. Do step 2 q-axis inductance 5.623 [mH]
and step 3 iteratively until 𝐿𝑑 ′ , 𝐿𝑞 ′ and 𝜓̂𝑃𝑀 converge to Stator resistant 0.086 [Ohm]
specific values (step 4). Magnet flux linkages 𝜓𝑃𝑀 0.105 [Wb]

Figure 6.Flow chart of proposed demagnetization fault diagnosis

A. Least Square Estimation of 𝜓̂𝑃𝑀 Figure 7.Torque, d- and q-axis current variation of IPMSM operation

We can get the degree of demagnetization in motor using TABLE II. COMPARISON OF INDUCTANCE IDENTIFICATION
Least Square method with rearranged equations of PMSM (3)
Torque [Nm] 2 4
~ (5). Estimate 𝜓̂𝑃𝑀 with normal motor will be used in
Inductance [mH] 𝐿𝑑 𝐿𝑞 𝐿𝑑 𝐿𝑞
updating inductance in next step.
Proposed method 2.026 5.972 1.967 5.895
B. Analytic Calculation of Inductance Conventional (RLS) 1.970 5.489 2.089 5.051
FEA result 2.083 6.165 2.028 5.623
Decreased flux density affects 𝜙𝑡𝑒𝑒𝑡ℎ,𝑃𝑀 and 𝜙𝑏𝑟𝑖𝑑𝑔𝑒,𝑃𝑀
in (17), (18). These changes make difference in (13), (14) and TABLE III. FLUX LINKAGES OF NORMAL AND 2 DEMAGNETIZED MOTORS
finally changes phase inductance with d- and q-axes aligned
Parameters normal Demagnetized 1 Demagnetized 2
(20), (22) and 𝐿𝑑 and 𝐿𝑞 .
𝜓𝑃𝑀 [Wb] 0.1055 0.0887 0.0729
C. Iteration: Estimation and Calculation Ratio (%) 100 84.1 69.1

Apply Least Square to estimate 𝜓̂𝑃𝑀 with updated 𝐿𝑑 and

The motor maintained 3000 rpm during the simulation with
𝐿𝑞 . Repeat Least Square estimation and analytic calculation
varying torque from 0 to 4 Nm (Fig. 7). We compared
until 𝐿𝑑 ′, 𝐿𝑞 ′ and 𝜓̂𝑃𝑀 converges. Finally we can get the inductance of normal motor when it is operated with 2 and 4
inductance and strength of flux linkages of demagnetized PM. Nm. As you can see in Table Ⅱ, analytic calculation of
inductance shows better result than conventional method
Proposed scheme could figure out variation of inductance which used online parameter estimation with Recursive Least
and flux linkages successfully with FEA simulation result and Square method (RLS) [11]. Proposed method shows errors
we will represent it in the next section. less than 5%. To diagnose demagnetization fault, the
accuracy of estimating 𝐿𝑑 is more important than that of 𝐿𝑞
V. SIMULATION RESULTS since in PMSM dynamics, terms including 𝐿𝑑 makes
significant role rather than terms including 𝐿𝑞 in equation (2).
IPMSM drive system was built in FEA program, Ansoft
We used two demagnetized motors which are realized by
Maxwell for motor implementation and Simplorer for eliminating PMs one by one so that the motors have reduced
inverter and control algorithm. Specification of the IPMSM is strength of PM. Magnet flux linkages of simulated
listed in Table Ⅰ and structure of the motor is given in Fig. 4. demagnetization fault IPMSM is given in Table Ⅲ.
We used Maximum Torque per Ampere (MTPA) to operate Result of inductance identification of demagnetized motors
IPMSM in simulation (Simplorer).
are respectively given in Table Ⅳ and Ⅴ. The parameters are

3189

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on October 14,2020 at 16:03:48 UTC from IEEE Xplore. Restrictions apply.
identified at 4 Nm operation. We could get more accurate which considers structural characteristics of the motor. Using
value of inductance than conventional estimation method in this technique, the variation of inductance was estimated very
both demagnetization cases (Table Ⅵ, Ⅶ). As we got more close to real value of FEA simulation. Applying Least Square
precise result in inductance, estimation of flux linkages also method with analytic calculated inductance showed accurate
had better performance than both using normal condition’s estimation result of flux linkages in demagnetized machines.
inductance with Least Square and RLS.
ACKNOWLEDGMENT
TABLE IV. ESTIMATED PARAMETERS OF DEMAGNETIZED IPMSM 1
"This research was supported by the MSIP (Ministry of
Proposed method Conventional (RLS) FEA Science, ICT and Future Planning), Korea, under the “IT
𝐿𝑑 2.196 1.597 2.162
Consilience Creative Program”
𝐿𝑞 6.077 5.285 5.844
(NIPA-2014-H0201-14-1001) supervised by the NIPA
(National IT Industry Promotion Agency)”
TABLE V. ESTIMATED PARAMETERS OF DEMAGNETIZED IPMSM 2

Proposed method Conventional (RLS) FEA REFERENCES

𝐿𝑑 2.262 2.801 2.264
𝐿𝑞 6.256 5.851 5.977 [1] Kwang Hee Nam, “AC Motor Control and Electric Vehicle
Applications”, CRC press, 2010, pp. 133.
[2] B. M. Ebrahimi, J. Faiz, and M. J. Roshtkhari, “Static-, Dynamic-, and
TABLE VI. FLUX LINKAGES ESTIMATION RESULT 1 Mixed-Eccentricity fault Diagnoses in Permanent-magnet Synchronous
Motor”, IEEE trans. Industrial electronics , vol. 56, no. 11, pp.
Proposed LS, normal RLS
FEA 4727-4739, Nov. 2009.
method 𝐿𝑑 , 𝐿𝑞 method [3] S. M. Cruz and A. Cardoso, “Stator fault diagnosis in three-phase
Flux linkages synchronous and asynchronous motors”, IEEE trans. Industrial
0.0875 0.0861 0.0849 0.0887
[Wb] Applications, vol. 37, no. 5, pp. 1227-1233, Sep. 2001.
Ratio of PM to [4] J. Rosero, L. Romeral, J.A. Ortega, J.C. Urresty, “Demagnetization
83.1 81.6 80.5 84.1
normal (%) fault detection by means of Hilbert Huang transform of the stator
Estimation current decomposition in PMSM”, IEEE International Symposium on
1.35 2.93 4.28 Industrial Electronics, pp. 172-177, June. 2008.
Error (%)
[5] Rosero, J. A, Cusido, J, Garcia, A, and Ortega, J. A, “Study on the
Permanent Magnet Demagnetization Fault in Permanent Magnet
TABLE VII. FLUX LINKAGE ESTIMATION RESULT 2
Synchronous Machines”, IEEE Industrial Electronics, IECON 2006,
Proposed LS, normal RLS pp. 879-884, Nov. 2006.
FEA [6] Riba Ruiz, J. –R, Rosero, J. A, Espinosa, A.G, and Romeral, L,
method 𝐿𝑑 , 𝐿𝑞 method
Flux linkages “Detection of Demagnetization faults in Permanent Magnet
0.0713 0.0677 0.0685 0.0729 Synchronous Motors under Nonstationary Conditions”, IEEE trans.
[Wb]
Magnetics, vol. 45, pp. 2961-2969, June 2009.
Ratio of PM to [7] Shyh-Jier Wang and Shir-Kuan Lin, “Analytical Prediction of the
67.6 64.2 64.9 69.1
normal (%) Incremental Inductance of the Permanent magnet Synchronous
Estimation Motors”, IEEE Trans. Magnetics, Vol. 40, No.$, July 2004.
2.19 7.13 6.04
Error (%) [8] K. J. Meessen, P. Thelin, J. Soulard, and E. A. Lomonova, “Inductance
Calculations of Permanent-magnet Synchronous Machines Including
As flux linkages of IPMSM gets weaker, reluctance of each Flux Change and Self- and Cross-Saturations”, IEEE Trans. Magnetics,
vol. 44, No. 10, October 2008
component in the motor get smaller so that the inductance [9] David Haliday, Robert Resnick and Jearl walker, Fundamentals of
gets bigger (Table Ⅳ, Ⅴ). Since inductance makes important Physics, 2004.
[10] Jewon Lee, Yong-Ju Jeon, Doo-chul Choi and SeungHun Kim,
role in motor dynamics (1), we need accurate value of “Demagnetization Fault Diagnosis Method for PMSM of Electric
inductance to estimate changes of flux linkages. Vehicle”, IEEE Industrial Electronics Society, IECON 2013, pp.
Using normal inductance in estimating flux linkages shows 2709-2713.
bigger error when demagnetization of PM gets worse (Table [11] Samul J. Underwood and Iqbal Husain, “Online Parameter Estimation
and Adaptive Control of Permanent-Magnet synchronous machines”,
Ⅵ, Ⅶ) since variation of inductance bigger. RLS method IEEE Trans. Industrial Electronics, vol. 57, no. 7, pp. 2435–2443, July
results better performance than formal technique but 2010.
proposed method showed even better performance than RLS.
Although we had poor result of estimating Lq , as mentioned
before, accuracy of Ld has significant role to estimating flux
linkages. So flux estimation showed better result and this
demonstrate that our approach can diagnose demagnetization
fault well.

VI. CONCLUSION
In this paper, PMSM demagnetization fault diagnosis
method subject to parameter variations has been proposed.
Proposed method is based on the analytic approach of PMSM

3190

Authorized licensed use limited to: Consortium - Algeria (CERIST). Downloaded on October 14,2020 at 16:03:48 UTC from IEEE Xplore. Restrictions apply.