optimization of bias current coefficient in the fault

13
Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=taut20 Automatika Journal for Control, Measurement, Electronics, Computing and Communications ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/taut20 Optimization of bias current coefficient in the fault-tolerance of active magnetic bearings based on the redundant structure parameters Xin Cheng, Shuai Deng, Baixin Cheng, Meiqian Lu & Rougang Zhou To cite this article: Xin Cheng, Shuai Deng, Baixin Cheng, Meiqian Lu & Rougang Zhou (2020) Optimization of bias current coefficient in the fault-tolerance of active magnetic bearings based on the redundant structure parameters, Automatika, 61:4, 602-613, DOI: 10.1080/00051144.2020.1806012 To link to this article: https://doi.org/10.1080/00051144.2020.1806012 © 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group Published online: 24 Aug 2020. Submit your article to this journal Article views: 248 View related articles View Crossmark data

Upload: others

Post on 16-Oct-2021

6 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Optimization of bias current coefficient in the fault

Full Terms & Conditions of access and use can be found athttps://www.tandfonline.com/action/journalInformation?journalCode=taut20

AutomatikaJournal for Control, Measurement, Electronics, Computing andCommunications

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/taut20

Optimization of bias current coefficient in thefault-tolerance of active magnetic bearings basedon the redundant structure parameters

Xin Cheng, Shuai Deng, Baixin Cheng, Meiqian Lu & Rougang Zhou

To cite this article: Xin Cheng, Shuai Deng, Baixin Cheng, Meiqian Lu & Rougang Zhou(2020) Optimization of bias current coefficient in the fault-tolerance of active magneticbearings based on the redundant structure parameters, Automatika, 61:4, 602-613, DOI:10.1080/00051144.2020.1806012

To link to this article: https://doi.org/10.1080/00051144.2020.1806012

© 2020 The Author(s). Published by InformaUK Limited, trading as Taylor & FrancisGroup

Published online: 24 Aug 2020.

Submit your article to this journal

Article views: 248

View related articles

View Crossmark data

Page 2: Optimization of bias current coefficient in the fault

AUTOMATIKA2020, VOL. 61, NO. 4, 602–613https://doi.org/10.1080/00051144.2020.1806012

REGULAR PAPER

Optimization of bias current coefficient in the fault-tolerance of activemagnetic bearings based on the redundant structure parameters

Xin Chenga,b, Shuai Denga, Baixin Chenga, Meiqian Lua and Rougang Zhouc

aSchool of Mechanical & Electronic Engineering, Wuhan University of Technology, Wuhan, People’s Republic of China; bHubei ProvincialEngineering Technology Research Center for Magnetic Suspension, Wuhan University of Technology, Wuhan, People’s Republic of China;cSchool of Mechanical Engineering, Hangzhou Dianzi University, Hangzhou, People’s Republic of China

ABSTRACTTo improve the reliability of magnetic bearings, the redundant structures are usually designedto provide the desired bearing characteristics continually by the reconfiguration of the remain-ing structures when some components fail. Bias current coefficient is one of the key coefficientsin the fault-tolerant control (FTC) of magnetically levitated bearings, and its inappropriate valuewill result in the failure of providing the desired bearing force due to saturation constraints. Thispaper presents optimization approaches of the bias current coefficient based on the redundantstructure parameters. By analysing the range of the bias current coefficient under saturation con-straints mathematically, the existence of the optimal solution has been proved, and the modelof the electromagnetic force (EMF), load current and the bias current coefficient has been estab-lished in this paper. In addition, algorithms to find the optimal solution has been designed fortwo kinds of optimization objectives, respectively, in FTC of magnetic bearings. Numerical veri-fications prove the effectiveness and the versatility of the proposed approaches in the differentstructures.

ARTICLE HISTORYReceived 18 June 2020Accepted 23 July 2020

KEYWORDSMagnetic bearings; biascurrent coefficient;optimization; fault-tolerance

1. Introduction

Magnetic bearings, support the rotors by electromag-netic force, with the advantages of no lubrication, nomechanical friction, and controllable support charac-teristics [1,2], are widely used in high-performanceequipment such as aero-engines, energy storage fly-wheels, nuclear turbine power generation equipment,etc., [1–3]. The basic control theory of magnetic bear-ings is based on the bias current linearization at theequilibrium position, namely, by the displacement-force coefficient and the current-force coefficient, elec-tromagnetic force is linearized [2].

The faults incorporated in magnetic bearings canbe divided into 3 types: controller, sensors and actua-tors. The hot standby configurations of the controlleror sensors are the common way to deal with the cor-responding faults in the magnetic bearing system [4,5],or more advanced hardware redundancy protocol forachieving high reliability, triple modular redundancy(TMR) [6,7], can be considered in some researches.However, there is special consideration of the faultsin the actuators. Because of the requirement of highefficiency in magnetic bearings, it is practically impos-sible to design some hot standby actuators preparedto replace the failed ones only. Moreover, there is asymmetry constraint on the stator structure [8], theshort circuit, break or partial insulation damages in

the electromagnetic coils will cause the unexpectedEMF, which is defined as the actuator fault [5] and willdestroy the original symmetry of the bearing structure,leading to the failure of bearing system and seriousimpacts [9].

Unlike hot standby configurations, another analyti-cal redundancy designwas introduced in [8,10] and caneffectively improve the reliability of magnetically levi-tated bearings. The failed actuators will be isolated andthe remaining parts will be continually used and recon-figured to support themagnetically levitated rotor. Sim-ilarly, the linearization of EMF generated by redun-dant supporting structure in magnetic bearings is animportant foundation for the realization of FTC, andrelevant researches basically follow this idea. Eric et al.[10] proposed the fault-tolerance of magnetic bearingsby generalized bias current linearization, in which, themagnetic flux lost due to failed actuator can be com-pensated by the current distribution, and a linear rela-tionship between the EMF and the controlled currentcan be derived. Na and Palazzolo [11] optimized theFTC of magnetic bearing and put forward an approachfor optimal selection of the current distribution matrixby Lagrange multiplier approach. Moreover, the cor-responding experimental verification was performedin their flexible rotor platform [12] to bear the rotorafter the failure of an actuator. Subsequently,Ming-Hsiu

CONTACT Rougang Zhou [email protected] School of Mechanical Engineering, Hangzhou Dianzi University, Hangzhou 310000, People’sRepublic of China

© 2020 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis GroupThis is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricteduse, distribution, and reproduction in any medium, provided the original work is properly cited.

Page 3: Optimization of bias current coefficient in the fault

AUTOMATIKA 603

Li [13] studied the magnetic circuit coupled actuatorstructure and extended the theory tomagnetic bearingsof composite structure with radial and axial structure.D. Noh et al. [14] experimentally verified the feasibil-ity of such bias current linearization theory based onthe molecular magnetic vacuum pump. In addition, inorder to effectively compensate for the model error ofthe linearization method, Meeker andMaslen [15] esti-mated factors such as magnetic leakage and edge andeddy current effects and established a more accuratemagnetic bearing model. Na and Palazzolo [16] cal-culated the reluctance of ferromagnetic material pathand modelling error due to magnetic leakage. The edgeeffects were replaced by means of a simple compensa-tion coefficient.

The aforementioned researches have already pre-sented a basic theoretical framework in the fault-tolerance of magnetic bearings, however, the follow-ing fields need to be concerned. (1) Analysis of thefault-tolerant control system (FTCS) model in mag-netic bearings. Arslan A-A. and KhalidM-H. presenteda comprehensive state-of-the-art review of FTCS withthe latest advances and applications in [17]. ActiveFTCS (AFTCS) consists of Fault Detection and Iso-lation (FDI) module [18], a reconfiguration mecha-nism and a reconfigurable controller [19,20]. Espe-cially, linear regression-based observer model can beused in the fault detection and isolation unit for faultdetection, isolation and reconfiguration in the AFTCS[21] to improve the system robustness. Passive FTCS(PFTCS) has no FDI unit and no controller reconfig-uration. Rather, the controller works in offline modein both normal and abnormal conditions with prede-fined parameters that mask the faulty readings from thecomponents [17]. Hybrid fault-tolerant control systempossesses properties of both active fault-tolerant con-trol system and passive fault-tolerant control system.A hybrid fault-tolerant control system was proposedin [22] for air–fuel ratio control of internal combus-tion gasoline engines based on Kalman filters and triplemodular redundancy. (2) Be similar with the commonmagnetic bearings, the ones with analytical redundancydesign have the requirements to linearize the EMFbased on a necessary bias design. In order to optimizethe EMF output in the FTC of magnetically levitatedbearings, Na and Palazzolo [23] designed an optimalselection scheme of current distribution matrix basedon the load current limitations and necessary lineariza-tion conditions, to lower the load current, comparedwith the current in [10]. Bias current coefficient is oneof the key coefficients in the general current lineariza-tion theory [11,14,24,25]. Na et al. [11] gets a certainoptimization result by selecting the central intensity ofmagnetic field as a bias current coefficient. Cheng X.et al. [25] established a FTC model, and through thenumerical verification, they pointed out that inappro-priate choice of bias current coefficient would cause the

failure of generating the desired EMF due to the satura-tion of the load current or the magnetic field. However,based on what we know, there is no further researchabout how to find the optimized bias current coefficientin FTCS for magnetic bearings.

Our contribution in this paper is an optimizationapproach of bias current coefficient based on the redun-dant supporting structure parameters. We define twokinds of optimization objectives, (1) maximum electro-magnetic force under the same structure and saturationconstraints and (2) minim intensity of magnetic fieldunder the same structure and EMFoutput. By analysingthe range of the bias current coefficient under satu-ration constraints mathematically, the existence of theoptimal solution has been proved, and the optimizationalgorithms have been designed to find the optimal solu-tion of bias current coefficient. Numerical verificationsprove the efficiency and the versatility of the proposedapproaches in different structures. The topic of thispaper is not the FTC algorithm but the optimizationof bias current coefficient in FTC. The implementa-tion and performance verification of FTC in magneticbearings can be found in [10–12,14,21].

Further contents of the paper are organized as fol-lows: Section 2 describes the structure parameters infault-tolerance of magnetic bearings. Section 3 pro-vides the mathematical proofs and Section 4 discussesthe optimization algorithms of the optimal bias cur-rent coefficient. The numerical verification is in Section5. The conclusion is presented in the last section withfuture directions.

2. Structure parameters in the fault-toleranceof magnetic bearing

Figure 1 illustrates the redundant structure and mag-netic circuit of radial magnetic bearing with n poles[10]. Relying on the coupled magnetic circuit betweenadjacent magnetic poles by the magnetic yoke, the lossofmagnetic flux caused by the failures of some actuatorscan be compensated to implement the fault-tolerance.The idea relieves the symmetry constraint on statorstructure.

The magnetic circuit equation is [10],

Rj�j − Rj+1�j+1 = NjIj − Nj+1Ij+1 (1)

where, Rj and �j are reluctance and magnetic flux ofthe jth pole, respectively; Nj and Ij are the number ofthe coil windings and the current of the jth pole, respec-tively. g(x, y)j,Aj andμ0 are the gap and area of jth pole,and vacuum magnetic permeability, respectively. Onecan write

Rj =g(x, y)jμ0Aj

(2)

Page 4: Optimization of bias current coefficient in the fault

604 X. CHENG ET AL.

Figure 1. Redundant structure (left), and magnetic circuit (right) of radial magnetic bearing.

Define

� = [�1 �2 · · · �j · · · �n

]T (3)

I = [I1 I2 · · · Ij · · · In

]T (4)

R =

⎡⎢⎢⎢⎢⎢⎢⎣

R1 −R2 0 · · · 0

0 R2 −R3. . .

...... 0

. . . . . . 00 · · · 0 Rn−1 Rn1 1 · · · 1 1

⎤⎥⎥⎥⎥⎥⎥⎦ (5)

N =

⎡⎢⎢⎢⎢⎢⎢⎣

N1 −N2 0 · · · 0

0 N2 −N3. . .

...... 0

. . . . . . 00 · · · 0 Nn−1 Nn0 0 · · · 0 0

⎤⎥⎥⎥⎥⎥⎥⎦ (6)

Then, the equation of coupling magnetic circuit is asfollowing [10]

RΦ = NI (7)

Note �j = BjAj, where Bj is the air gap magnetic fieldintensity of the jth pole. We define Aj = A, and

B = [B1 B2 · · · Bj · · · Bn

]T (8)

B can be described as

B = A−1R−1NI (9)

In the Equation (9),A represents the diagonalmatrixof the magnetic pole area. Fx and Fy are the resultantforces of the EMF in the x and y directions, respectively.Then one can write⎧⎪⎪⎪⎪⎪⎪⎨

⎪⎪⎪⎪⎪⎪⎩

Fx = A2μ0

BTDxB

Fy = A2μ0

BTDyB

Dx = diag[cos θ1 cos θ2 · · · cos θn

]Dy = diag

[sin θ1 sin θ2 · · · sin θn

](10)

It can be seen that even if an electromagnetic coilfails, the corresponding magneto-motive force is 0, butthe Equation (10) can still hold, namely, the recon-figuration of the supporting force is realized by thecompensation of magnetic flux. We define

V = A−1R−1N (11)

The diagonal matrix K is introduced to describethe state of the electromagnetic coil. If a coil fails, thecorresponding diagonal element is 0.

K = diag[1 1 · · · 1

]︸ ︷︷ ︸n

(12)

Then, Fx and Fy can be described as [10]{Fx = A

2μ0ITKTVTDxVKI

Fy = A2μ0

ITKTVTDyVKI(13)

Current vector is defined as IC = [C0 ix iy

]T , whereC0 is the bias current coefficient, ix and iy are the controlcurrents in the x and y directions, respectively. CurrentdistributionmatrixW is defined in [10], which satisfiesI = WIC, and⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨

⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩

WTKTVTDxVKW = Mx

WTKTVTDyVKW = My

Mx =

⎡⎢⎣ 0 0.5 00.5 0 00 0 0

⎤⎥⎦

My =

⎡⎢⎣ 0 0 0.5

0 0 00.5 0 0

⎤⎥⎦

(14)

So the Equation (14) can be simplified to⎧⎨⎩Fx = A

2μ0C0ix

Fy = A2μ0

C0iy(15)

Page 5: Optimization of bias current coefficient in the fault

AUTOMATIKA 605

Usually, C0 can be set as a constant to decouple andlinearize the relationship of EMF and currents. How-ever, there are constraints in magnetic bearings sys-tem, e.g. the magnetic field saturation due to magneticmaterial or the current saturation of power amplifier.Considering that inappropriate value of C0 may leadto failure of outputting desired EMF due to the aboveconstraints, how to optimize the bias current coeffi-cient should be taken into consideration. We definetwo kinds of optimization objectives, (1) maximumelectromagnetic force under the same structure andsaturation constraints, and (2) minimum intensity ofmagnetic field under the same structure and EMFoutput.

3. Mathematical proofs of the optimal biascurrent coefficient

Assuming the area of each magnetic pole A, the num-ber of turns of the coil N and the air gaps g0 are thesame, then we introduce them into Equations. (2), (5),(6), (11), can get Equation (16).

V = u0Nng0

⎡⎢⎢⎢⎢⎣n − 1 −1 · · · −1

−1 n − 1. . .

......

. . . . . . −1−1 · · · −1 n − 1

⎤⎥⎥⎥⎥⎦n×n

(16)

Therefore, we can have⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩

VTDxV = N2μ0A2n2g20

L

· diag[cos θ1 cos θ2 · · · cos θn] · LVTDyV = N2μ0A

2n2g20L

· diag[sin θ1 sin θ2 · · · sin θn] · L

L =

⎡⎢⎢⎢⎢⎢⎣n − 1 −1 · · · −1

−1 n − 1. . .

......

. . . . . . −1−1 · · · −1 n − 1

⎤⎥⎥⎥⎥⎥⎦n×n

(17)

Define ⎧⎨⎩Px = 2g20

N2μ0AVTDxV

Py = 2g20N2μ0A

VTDyV(18)

and introduce Equations (17) and (18) into Equation(14) to get Equation (19)⎧⎨

⎩WTKTPxKW = 2g20N2μ0A

Mx

WTKTPyKW = 2g20N2μ0A

My

(19)

The solution of the current-distribution matrix canbe divided into two parts, one part is the solutionmatrixWn×3, which is only related to the number of magnetic

poles and the corresponding angle of each magneticpole; the other part is a scalar solution which is relatedto the structure parameters A,N,g0 and μ0 in Equation(20).

W =√2g0

N√

μ0AWn×3 (20)

From Equation (19), it can be seen that for a cer-tain redundant structure, its parameters will become apart of the current-distribution matrix W by means ofconstant. However, Wdetermines the performance ofthe fault-tolerant control together with the bias currentcoefficient. Introduce Equation (15) into I = WIC toget Equation (21).

I = W ·⎡⎣ C0Fx/C0Fy/C0

⎤⎦ (21)

Define the general form of current-distributionmatrix as in Equation (22), the current expressionas Equation (23) can be derived by Equations (21)and (22).

W =

⎡⎢⎢⎢⎣a11 a12 a13a21 a22 a23...

......

an1 an2 an3

⎤⎥⎥⎥⎦ (22)

I =

⎡⎢⎢⎢⎢⎢⎣a11C0 + a12 · Fx

C0+ a13 · Fy

C0

a21C0 + a22 · FxC0

+ a23 · FyC0

...an1C0 + an2 · Fx

C0+ an3 · Fy

C0

⎤⎥⎥⎥⎥⎥⎦ (23)

The ith pole current can be expressed as

Ii = ai1C0 + ai2 · FxC0

+ ai3 · FyC0

0 < i ≤ n (24)

As shown in Figure 2, we define

�F = F∠β = Fx · �i + Fy · �j (25)

Then Equation (24) can be expressed as

Ii = ai1C0 + FC0

· (ai2 cosβ + ai3 sinβ) 0 < i ≤ n

(26)We define {

Ki1 = ai1Ki2 = ai2 cosβ + ai3 sinβ

(27)

Equation (24) can be simplified as

Ii = Ki1C0 + Ki2FC0

0 < i ≤ n (28)

Usually, W is obtained offline and stored in theFTCS, which means that Kij is a constant for the any

Page 6: Optimization of bias current coefficient in the fault

606 X. CHENG ET AL.

Figure 2. Exploitation of the desired force vector.

identifiedW. Therefore, it can be known fromEquation(28) that the value of C0 directly affects the relationshipbetween the desired EMF and the controlled current ofeach magnetic pole. the inappropriate value of C0 willresult in the inability to output the desired EMFbecauseof saturation constraint. Moreover, there is an optimalvalue of C0 in the any of following four conditions inFigure 3 from Equation (28), and the 4 conditions arefrom the combinations of K i1 and K i2 in Equation (27).

Obviously, the above ranges are symmetrical. Inorder to simplify the model, we only consider the pos-itive value of C0. Therefore the desired EMF can begenerated only when C0 ∈ (C3,C4), where⎧⎪⎨

⎪⎩C3 =

∣∣∣Imax−√

I2max−4Ki1Ki2·F∣∣∣

2|Ki1|

C4 = Imax+√

I2max−4Ki1Ki2·F2|Ki1|

(29)

4. Optimization algorithms

4.1. Case A: Maximum electromagnetic forceoutput under the same saturation constraints

Convert Equation (26) to Equation (30),(C0 − Ii

2ai1

)2+ F

(ai2 cosβ + ai3 sinβ)

ai1= Ii2

4ai12(30)

For any identifiedW and β , when Ii takes the max-imum value, namely Imax, an optimal value of bias cur-rent coefficient can be got as in Equation (31), so that

Figure 3. The ranges of C0 under current saturation constraints. (a) Ki1 > 0, Ki2 > 0, (b) Ki1 > 0, Ki2 < 0, (c) Ki1 < 0, Ki2 > 0 and (d).Ki1 < 0, Ki2 < 0.

Figure 4. Current of every magnetic pole. (a) A magnetic pole current exceeds Imax. (b) Two magnetic pole currents reach Imax.

Page 7: Optimization of bias current coefficient in the fault

AUTOMATIKA 607

Figure 5. Flow chart of optimization algorithm for case A.

the corresponding structure can provide the maximumEMF.

C0 = Ii2ai1

(31)

However, it is necessary to confirmwhether the opti-mal C0 satisfying Equation (31) exists within the rangein Equation (29); if not, namely, when any magneticpole current reaches the Imax, there must be some othermagnetic pole current exceeding the Imax, as in Figure4(a). This case means we cannot get the optimal C0fromEquation (31), and need to consider that twomag-netic pole currents reach the Imax at the same time, asin Figure 4(b). Assuming the q th and the p th magneticpole current reach the Imax at the same time, the currentexpression is from Equation (32).

⎧⎨⎩Imax = Kp1C0 + Kp2F

C0

Imax = Kq1C0 + Kq2FC0

(32)

The optimal value of C0can be got from Equation (33).

⎧⎪⎨⎪⎩C0 = (Kq2−Kp2)Imax

Kp1Kq2−Kq1Kp2

F = (Kp1−Kq1)(Kq2−Kp2)I2max(Kp1Kq2−Kq1Kp2)

2

(33)

Figure 5 illustrates the flow chart of optimizationalgorithm for case A.

4.2. Case B: Minimum intensity ofmagnetic fieldunder the same electromagnetic force output

From Equations (9), (11) and (21), the followingEquation (34) can be obtained.

B = V · W ·⎡⎣ C0Fx/C0Fy/C0

⎤⎦ (34)

Since the matrix V is determined by the structureparameters, we defineW′ = V · W, can have

B = W′ ·⎡⎣ C0Fx/C0Fy/C0

⎤⎦ (35)

Combined with Equation (28), we can have,

Bi = Ki1C0 + Ki2FC0

0 < i ≤ n (36)

The relationship between the intensity of magneticfield and C0 of each magnetic pole can be accuratelyobtained according to the matrix V and W. Once thedesired output EMF is generated, we need to find theoptimal C0 with the minimal value of the maximumintensity of magnetic field of all the poles, comparedwith other value of C0. There are two cases, as shownin Figure 6.

In Figure 6(a), for the lowest point of the kth polecurve, when

C0 =√∣∣∣∣Kk2F

Kk1

∣∣∣∣ (37)

themagnetic flux intensity of a singlemagnetic pole canbe taken to the minimum value of Bk = 2

√Kk1Kk2F or

Bk = 0. In Figure 7(b), the lowest point is the inter-section of the two curves, therefore the intensity ofmagnetic field expression corresponding to the sth andrth magnetic poles can be obtained from the Equation

Figure 6. Intensity of magnetic field vs C0 for all the poles. (a) The lowest point of maximum intensity of magnetic field is not theintersection of two curves. (b) The lowest point is the intersection of two curves.

Page 8: Optimization of bias current coefficient in the fault

608 X. CHENG ET AL.

Figure 7. Flow chart of optimization algorithm for case B.

Figure 8. Octupole radial redundant structure.

(39).

B =∣∣∣∣Ks1C0 + Ks2F

C0

∣∣∣∣ =∣∣∣∣Kr1C0 + Kr2F

C0

∣∣∣∣ (38)

The solution of Equation (38) is⎧⎪⎪⎨⎪⎪⎩C0 =

√∣∣∣ (Kr2−Ks2)F(Ks1−Kr1)

∣∣∣B = |Ks1Kr2 − Kr1Ks2|

√∣∣∣ F(Ks1−Kr1)(Kr2−Ks2)

∣∣∣(39)

Flow chart of optimization algorithm for case B is inFigure 7, it is obvious that the optimal C0 can always begot from one of the Equations (31), (33), (37) and (39).

5. Numerical verification

By taking an 8-pole symmetrical radial magnetic bear-ing as an example, its structure is shown in Figure 8 andparameters are presented in Table 1.

Based on Equation (5), we can have Equation (40)for this example.

R = g0u0A

·

⎡⎢⎢⎢⎢⎢⎢⎣

1 −1 0 · · · 0

0 1 −1. . .

...... 0

. . . . . . 00 · · · 0 1 −11 1 · · · 1 1

⎤⎥⎥⎥⎥⎥⎥⎦

8 × 8

(40)

By introducing the structure parameters intoEquation (18), we can have

⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩

Px =√2

16 ·

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

6√2 −1 − √

2 −√2 1 − √

2 0 1 − √2 −√

2 −1 − √2

−1 − √2 6 −1 0

√2 − 1 0 −1 −2

−√2 −1 0 1

√2 1 0 −1

1 − √2 0 1 −6 1 + √

2 2 1 00

√2 − 1

√2 1 + √

2 −6√2 1 + √

2√2

√2 − 1

1 − √2 0 1 2 1 + √

2 −6 1 0−√

2 −1 0 1√2 1 0 −1

−1 − √2 −2 −1 0

√2 − 1 0 −1 6

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦

Py =√2

16 ·

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

0 −1 −√2 −1 0 1

√2 1

−1 6 −1 − √2 −2 −1 0

√2 − 1 0

−√2 −1 − √

2 6√2 −1 − √

2 −√2 1 − √

2 0 1 − √2

−1 −2 −1 − √2 6 −1 0

√2 − 1 0

0 −1 −√2 −1 0 1

√2 1

1 0 1 − √2 0 1 −6 1 + √

2 2√2

√2 − 1 0

√2 − 1

√2 1 + √

2 −6√2 1 + √

21 0 1 − √

2 0 1 2 1 + √2 −6

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦

(41)

Page 9: Optimization of bias current coefficient in the fault

AUTOMATIKA 609

Table 1. Parameters of the structure in Figure 8.

Parameter Value Unit

Number of circles, N 200 –Air gap, g0 10−4 mSaturation current, Imax 5 APole angle, θ 45 degreePole area, A 491 mm2

Intensity of magnetic saturation, Bmax 1.2 T

Table 2. Computational cost ofthe proposed algorithm.

Case Value Unit

A 6.6 usB 52.8 us

Wecan find one solution of W in Equation (42) from[10], with identity matrix K . Obviously, the structureparameters of themagnetic bearing are contained inW,as constants.

W = g04N

√μ0A

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

2 2 0−2 −√

2 −√2

2 0 2−2

√2 −√

22 −2 0

−2√2

√2

2 0 −2−2 −√

2√2

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦

(42)

The computational cost of the proposed algorithmof different cases is shown in Table 2; the values aremeasured by CCS Profile clock when the proposedalgorithm is running in the TMS320F28335 DSP withthe core clock of 150MHz.

5.1. Maximum EMF output

Firstly, we consider the condition of no failed actua-tor. Figure 9(a) illustrates the maximum output EMFunder variable Imax; Figure 9(b) describes the relation-ship between the C0 and the maximum EMF outputwhen Imax is 5A.we can find the optimalC0 in Equation(43) based on the proposed approach.{

C0 = I1/2a11 = 24.8397F = 617.0088

(43)

Figures 10–11 illustrates the same relationshipswhen the 8th actuator fails, and theW for this conditionis shown in Equation (44).

W =

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣

0.2013 0.1718 −0.07120 0 −0.1423

0.2013 0.0712 0.02950 0.1423 −0.1423

0.2013 −0.0295 −0.07120 0.1423 0

0.2013 0.0712 0.17180 0 0

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦

(44)

5.2. Minimum intensity ofmagnetic field

When expected EMF is[Fx Fy

] = [500 N 0 N

],

we can find the optimal solution C0, consideringtwo same failure condition based on the proposedapproach, (1) no failed actuator, C0 = 22.3607, andBmax = 1.13 T; (2) the 8th actuator fails, C0 = 22.3593,and Bmax = 1.13 T.

5.3. Performance simulations of FTC adoptingproposed optimization algorithms

The proposed optimization algorithms can help to findthe optimal C0 in the FTC of magnetic bearing. Theimplementation details of FTC of magnetic bearingscan be found in [10–12,14,21]. To verify the effective-ness of the proposed optimization approach, the simu-lations based on Matlab/Simulink was carried on. Thestructure and parameters of magnetic bearing are inFigure 8 and Table 1, respectively.

As in Figure 12(a), curve B demonstrates the trajec-tory of rotor when the optimized C0 = 23 is chosen.The failure conditions were designed as follows. (1) at0.5 s, the rotor was suspended near its equilibriumposi-tion (0.001m); (2) the 8th coil failed at 1 s; (3) thenthe 6th coil failed at 1.5 s. It is obvious that the canreturn to the equilibrium position after the failures ofcoils because the supporting structure was reconfiguredunder the FTC. Figure 12(b) describes the electromag-netic force in the reconfiguration, it is clear that theoutput electromagnetic force (curve D) can compen-sate the disturbances in the reconfiguration tomaintainthe resultant force (curve C) near the 0N. The trajec-tories of rotor (curve A and B) are similar if ignoringthe reconfiguration, it means that whether or not theoptimal C0 is chosen, the performance of FTC in thiscase is similar to effectively deal with the failure ofcoils.

Figure 13 demonstrates different performances if asinusoidal disturbance with the magnitude of ± 300Nand frequency of 100 rad/s is added into the controlloop. The optimized C0 chosen in FTC can effectivelycompensate the disturbance(in curve B) comparedwiththe case in curve A, it means that the optimized C0can effectively add the EMF output capacity if theextreme cases occur. Moreover, the real-time perfor-mance of FTC will have the decisive effects to maintainthe stability of rotor in the reconfiguration if some coilsfail.

5.4. Versatility

To prove the versatility of proposed approach, a differ-ent redundant structure in Figure 14 is chosen, whilethe parameters are in Table 3. A matrix W for this

Page 10: Optimization of bias current coefficient in the fault

610 X. CHENG ET AL.

Figure 9. (a) Maximum EMF vs C0 when Imax is 5 A. (b) Maximum EMF vs Imax.

Figure 10. When the 8th actuator fails, the curves of (a) Maximum EMF vs C0, and (b) Maximum EMF vs Imax.

Figure 11. Bmax vs C0 under the condition of (a) no failed actuator, and (b) the 8th actuator fails.

structure can be got in Equation (45).

W =

⎡⎢⎢⎢⎢⎢⎢⎣

0.3932 0.2241 0−0.3932 −0.1691 −0.22410.3932 −0.1691 0.2241

−0.3932 0.2241 00.3932 −0.1691 −0.2241

−0.3932 −0.1691 0.2241

⎤⎥⎥⎥⎥⎥⎥⎦ (45)

Figures 15 and 16 illustrate the similar results as in Fig-ures 10 and 11 for the 6-pole structure in Figure 14.It can be seen that the optimal C0 can be got basedon the discussed approach even for this 6-pole struc-ture, which means the good versatility of the discussedapproach.

Page 11: Optimization of bias current coefficient in the fault

AUTOMATIKA 611

Figure 12. Simulation results of FTC, (a) trajectory of rotor, A-a random selection, C0 = 7, B-the optimized C0 = 23. (b) EMF in thereconfiguration, C-resultant force, D-the output EMF.

Figure 13. Simulation results of FTC. Trajectory of rotor in the reconfiguration time of 1ms (a) and 10ms (b), under a sinusoidaldisturbance, the magnitude of± 300 N and frequency of 100 rad/s. A-a random selection, C0 = 7, B-the optimized C0 = 23.

Figure 14. Another 6-pole redundant supporting structure.

6. Conclusions

Aiming at the FTC of active magnetic bearings, thispaper proposes optimization approaches of bias current

Table 3. Parameters of the structure in Figure. 14.

Parameter Value Unit

Number of circles, N 85 –Air gap, g0 4× 10−4 mSaturation current, Imax 5 APole angle, θ 60 degreePole area, A 57 mm2

Intensity of magnetic field saturation, Bmax 1.2 T

coefficient based on structure parameters. The relevantconclusions are as follows,

(1) For a certain structure, there is an optimal C0, andonly when the bias current coefficient is within acertain range, the desired EMF can be generated.This paper proves the existence of the range.

(2) To the two kinds of optimization objectives, (a)maximum EMF under the same structure and

Page 12: Optimization of bias current coefficient in the fault

612 X. CHENG ET AL.

Figure 15. (a) Maximum electromagnetic force vs C0. (b) Maximum electromagnetic force vs Imax, for the structure in Figure 14.

Figure 16. Bmax vs C0 for the structure in Figure 14.

saturation constraints and (b) minim intensityof magnetic field under the same structure andEMFoutput, we design effective optimization algo-rithms to find the optimal solution of C0. In addi-tion, the versatility of proposed approaches hasbeen proved.

The limitation of the proposed approach is thatthe basic electromagnetic force linearization theoryis based on the assumption, namely, the solution ofEquation (14) must consider the constraints in theactual applications, as follows: (1) the saturation ofmagnetic field; (2) the saturation of current in thepower amplifier; (3) the redundancy of the magneticbearings – the more redundant coils, the more failedones allowed; (4) the ratio of maxim EMF required andEMF designed; (5) the rotor dynamics required or pre-cision; (6) the failed coils are continuously or discretelyarranged.

The future work should focus on the optimizationapproach of bias current coefficient under certain con-straints, and the size of the redundancy required forcertain types of faults.

Acknowledgements

This work was supported National Natural Science Funds ofChina under Grants 51575411, the National Key Program

of China under Grant No. 2018YFB2000103 and the Funda-mental Research Funds for the Central Universities of Chinaunder Grants 2020-YB-022 and WUT2019III143CG.

Methodology, Xin Cheng; Validation, Meiqian Lu andShuai Deng; Writing – original draft preparation, MeiqianLu; Writing – review and editing, Baixin Cheng; supervision,Rougang Zhou.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This work was supported National Natural Science Fundsof China under Grants [51575411], the National KeyProgram of China under [Grant No. 2018YFB2000103]and the Fundamental Research Funds for the CentralUniversities of China under Grants [2020-YB-022 andWUT2019III143CG].

References

[1] Aenis M, Knopf E, Nordmann R. Active magnetic bear-ings for the identification and fault diagnosis in turbo-machinery. Mechatronics. 2002;12(8):1011–1021.

[2] Cheng X, Zhang L, Zhou RG, et al. Analysis ofoutput precision characteristics of digital switchingpower amplifier in active magnetic bearing system.Automatika. 2017;58(2):205–215.

[3] Aanstoos TA, Kajs JP, BrinkmanWG, et al. HighVoltageStator for a Flywheel Energy Storage System. IEEETransMagn. 2001;37(1):242–247.

[4] Cheng X, Wang B, Chen Q, et al. A unified designand the current ripple characteristic analysis of digitalswitching power amplifier in active magnetic-levitatedbearings. Int J Appl Elect Mech. 2017;55:391–407.

[5] Schweitzer G, Maslen EH. Magnetic bearings: theory,design, and application to rotating machinery. Berlin:Springer; 2009. ch. 14, pp. 407–433.

[6] Arslan AA, Khalid M-H. Advanced fault tolerant air-fuel ratio control of internal combustion gas enginefor sensor and actuator faults. IEEE Access. 2019;7:17634–17643.

[7] Hudson S, Sundar RSS, Koppu S. Fault control usingtriple modular redundancy (TMR), in progress in com-puting, analytics and networking. Singapore: Springer;2018. pp. 471–480.

Page 13: Optimization of bias current coefficient in the fault

AUTOMATIKA 613

[8] Cheng X, Chen Q, Zeng H, et al. Reconfiguration rulesfor loosely-coupled redundant supporting structure inradial magnetic bearings. Int J Appl Elect Mechan.2016;51(2):91–106.

[9] Cheng X, Cheng B, Lu M, et al. An online fault-diagnosis of electromagnetic actuator based on vari-ation characteristics of load current. Automatika.2020;61:11–20.

[10] Maslen EH, Meeker DC. Fault tolerance of magneticbearings by generalized bias current linearization. IEEETrans Magn. 1995;31(3):2304–2314.

[11] Na UJ, Palazzolo AB. Optimized realization of fault-tolerant heteropolar magnetic bearings. J Vib Acoust-trans ASME. 2000;122:209–221.

[12] Na UJ, Palazzolo AB, Provenza A. Test and the-ory correlation study for a flexible rotor on fault-tolerant magnetic bearings. J VibAcoust, Trans ASME.2002;124:359–366.

[13] Li MH, Palazzolo A, Kennyetal A. Fault-toleranthomopolar magnetic bearings. IEEE Trans Magn.2004;40(5):3308–3318.

[14] NohMD, Cho S-R, Kyung J-H, et al. Design and imple-mentation of a fault-tolerant magnetic bearing sys-tem for Turbo-Molecular Vacuum Pump. IEEE-ASMETrans Mechatron. 2005;10:626–631.

[15] Meeker DC, Maslen EH, Noh MD. An augmentedcircuit model for magnetic bearings including eddycurrents, fringing and leakage. IEEE Trans Magn.1996;32:3219–3227.

[16] Na UJ, Palazzolo AB. Fault tolerance of magnetic bear-ings withmaterial path reluctances and fringing factors.IEEE Trans Magn. 2000;36:3939–3946.

[17] Arslan AA, Khalid M-H. A review of fault tolerant con-trol systems: advancements and applications. Measure-ment. 2019;143:58–68.

[18] Alwi H, Edwards C, Tan CP. Fault tolerant control andfault detection and isolation. Fault Detect. Fault-Toler.Control Using SlidingModes, Springer-Verlag, London,2011, 7–27.

[19] Wang Y, Donghua Zhou S, Qin J, et al. Active fault-tolerant control for a class of nonlinear systems withsensor faults. Int J Control Autom Syst. 2008;6:339–350.

[20] Yang F, Zhang H, Jiang B, et al. Adaptive reconfigurablecontrol of systems with time-varying delay againstunknown actuator faults. Int J Adapt Control SignalProcess. 2014;8:1206–1226. doi:10.1002/acs.2439.

[21] Arslan AA, Khalid M-H. Hybrid fault tolerant controlfor air–fuel ratio control of internal combustion gaso-line engine using Kalman filters with advanced redun-dancy. Meas control. 2019;52(5–6):473–492.

[22] Arslan AA, Khalid M-H. Robust active fault-tolerantcontrol for internal combustion gas engine for air–fuelratio control with statistical regression-based observermodel. Meas Control. 2019;52:1179–1194.

[23] Na UJ, Palazzolo AB. The fault-tolerant control of mag-netic bearings with reduced controller outputs. J DynSyst, Meas Control-trans ASME. 2001;123:219–224.

[24] Cheng X, Cheng BX, Liu H, et al. An accurate lineariza-tion of electromagnetic force of heteropolar magneticbearings with redundant structures. J Eng Gas TurbinesPower, Trans ASME. 2020. DOI:10.1115/1.4046703

[25] Cheng X, Liu H, Song S, et al. Reconfigurationof tightly-coupled redundant supporting structure inactive magnetic bearings under the failures of elec-tromagnetic actuators. Int J Appl Electromagn Mech.2017;54:421–432.