(19)
(11)EP 3 943 768 B1

(12)EUROPEAN PATENT SPECIFICATION

(45)Mention of the grant of the patent:
07.09.2022 Bulletin 2022/36

(21)Application number: 20886182.3

(22)Date of filing:  24.07.2020
(51)International Patent Classification (IPC): 
F16C 32/04(2006.01)
(52)Cooperative Patent Classification (CPC):
F16C 32/0451; F16C 32/0453; H02P 25/06; H02P 23/0004; F16C 32/0455; F16C 2360/45; F16C 2360/31; F16C 2300/22
(86)International application number:
PCT/CN2020/103958
(87)International publication number:
WO 2021/237910 (02.12.2021 Gazette  2021/48)

(54)

ACTIVE MAGNETIC BEARING CONTROLLER CONSTRUCTION METHOD BASED ON TABLE LOOKUP METHOD

VERFAHREN ZUR KONSTRUKTION EINER AKTIVEN MAGNETISCHEN LAGERSTEUERUNG BASIEREND AUF EINEM TABELLENSUCHVERFAHREN

PROCÉDÉ DE CONSTRUCTION DE DISPOSITIF DE COMMANDE DE PALIER MAGNÉTIQUE ACTIF BASÉ SUR UN PROCÉDÉ DE CONSULTATION DE TABLE


(84)Designated Contracting States:
AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

(30)Priority: 29.05.2020 CN 202010473087

(43)Date of publication of application:
26.01.2022 Bulletin 2022/04

(73)Proprietor: Jiangsu University
Zhenjiang, Jiangsu 212013 (CN)

(72)Inventors:
  • SUN, Xiaodong
    Zhenjiang Jiangsu 212013 (CN)
  • JIN, Zhijia
    Zhenjiang Jiangsu 212013 (CN)
  • CHEN, Long
    Zhenjiang Jiangsu 212013 (CN)
  • YANG, Zebin
    Zhenjiang Jiangsu 212013 (CN)
  • ZHOU, Weiqi
    Zhenjiang Jiangsu 212013 (CN)
  • LI, Ke
    Zhenjiang Jiangsu 212013 (CN)

(74)Representative: Laufhütte, Dieter 
Lorenz Seidler Gossel Rechtsanwälte Patentanwälte Partnerschaft mbB Widenmayerstraße 23
80538 München
80538 München (DE)


(56)References cited: : 
WO-A1-99/40334
CN-A- 103 645 637
CN-A- 110 552 961
CN-B- 103 076 740
CN-A- 103 034 127
CN-A- 108 909 524
CN-A- 110 552 961
  
      
    Note: Within nine months from the publication of the mention of the grant of the European patent, any person may give notice to the European Patent Office of opposition to the European patent granted. Notice of opposition shall be filed in a written reasoned statement. It shall not be deemed to have been filed until the opposition fee has been paid. (Art. 99(1) European Patent Convention).


    Description

    Technical Field



    [0001] The present invention relates to a technology of controlling active magnetic bearings, which is applicable to control of active magnetic bearings in high-speed operating devices such as high-speed compressors, wind power generators, and molecular pumps and belongs to the technical field of magnetic suspension, and in particular, to a method for constructing an active magnetic bearing controller.

    Background



    [0002] A magnetic bearing is a rotor support system that uses the electromagnetic force to overcome the gravity and the interference force of a rotor and realizes suspension of the rotor without mechanical contact. It has the characteristics of no mechanical contact, long service life, and easy maintenance. Besides, its stiffness and damping are adjustable, and the output of the electromagnetic force can be flexibly adjusted by controlling the current in the windings to realize dynamic adjustment of the stiffness and damping of the magnetic bearing. At present, applications of molecular pumps, wind power generation, flywheel energy storage, and the like that use high-speed rotating shafts are increasing. Conventional mechanical bearings will greatly reduce the service life of the devices due to friction loss. Therefore, the applications of magnetic bearings have been continuously promoted.

    [0003] Magnetic bearings can be divided into three types: active, passive, and hybrid magnetic bearings. The active magnetic bearings are widely used due to advantages such as simple structures and adjustable suspension forces. Active magnetic bearings are for example known from CN 103 076 740 B. However, at present, the stability control of the active magnetic bearings has defects, and the biggest problem lies in the inability to obtain accurate control models, that is, current stiffness and displacement stiffness. The existing control methods generally use fixed current stiffness coefficients and displacement stiffness coefficients. However, the displacements, currents, and suspension forces of the active magnetic bearings are in approximately linear relationships only when the active magnetic bearings are at working points. When the displacements and the currents change greatly, the models will no longer be accurate. Fuzzy logic control, neural network control, parameter adjustment control based on advanced algorithms, and the like are generally used in the prior art. They make very limited adjustments to the models, or their adjustment of control only lies in the adjustment of controller parameters. Thus, there is an urgent need for a reasonable method that can change model parameters with the displacement change of the rotor of the active magnetic bearing to achieve more accurate control.

    [0004] At present, the look-up table method has been used to a certain extent in the control methods of various fields. The look-up table method is a control method which obtains corresponding results through data in a table or through interpolation calculation by using known test or simulation data in the control. It is also widely used in the electrical field. For example, in the control of a switched reluctance motor, flux linkage information is obtained through look-up table by using position information from a sensor and the magnitude of a control current. Therefore, with reference to the existing look-up table method, it is a feasible solution to obtain accurate model parameters by using the look-up table method in the control of the active magnetic bearings. However, the existing look-up table method has some problems. The major problem is that the creation of a usable parameter table requires plenty of experiments or simulations and repeated verifications, so that the experimental costs and time costs are high. Therefore, how to quickly create a parameter table with high accuracy is a problem worth studying.

    [0005] The Kriging model is a regression algorithm for spatial modeling and prediction of a random process or random field based on a covariance function. In a specific random process, for example, an intrinsically stationary process, the Kriging model can provide the best linear unbiased estimation (BLUE), which is also called spatial BLUE in statistics. Therefore, the Kriging model is applied in many fields such as geographical science, environmental science, and atmospheric science. In many cases, non-stationary regionalized variables occur, and a universal Kriging model is required for processing. Through the universal Kriging model, a parameter table of an active magnetic bearing can be obtained by using only a small amount of data.

    Summary



    [0006] The objective of the present invention is to provide a method for constructing an active magnetic bearing controller based on a look-up table method, to solve the problem of inaccurate models in the control of active magnetic bearings in the prior art. In the method, universal Kriging models are used to create model state tables, the control of an active magnetic bearing is implemented through the look-up table method, and current stiffness coefficients and displacement stiffness coefficients can be adjusted in real time according to displacement and current changes of the active magnetic bearing during operation.

    [0007] The method for constructing an active magnetic bearing controller based on a look-up table method according to the present invention adopts a technical solution with the following steps:

    Step (1): building finite element models of an active magnetic bearing, and obtaining, by using the finite element models and based on a general universal Kriging model, two universal Kriging prediction models about actual suspension forces x, y being in association with actual displacement eccentricities 0, 0 and actual control currents x, y in X-axis and Y-axis directions;

    Step (2): creating, based on the two universal Kriging prediction models, two model state tables about the actual suspension forces x, y being in association with the actual displacement eccentricities 0, 0 and the actual control currents x, y to construct two corresponding look-up table modules with the two model state tables being built in, respectively; and

    Step (3): constructing an active magnetic bearing controller by using two fuzzy adaptive PID controllers, two amplifier modules, two look-up table modules, and two measurement modules in the X-axis and Y-axis directions, wherein the fuzzy adaptive PID controller, the amplifier module, and the look-up table module in the X-axis direction are connected in series and then connected to an input end of the active magnetic bearing; the fuzzy adaptive PID controller, the amplifier module, and the look-up table module in the Y-axis direction are connected in series and then connected to the input end of the active magnetic bearing; the two measurement modules in the X-axis and Y-axis directions measure the actual displacement eccentricities 0, 0 of the active magnetic bearing in the X-axis and Y-axis directions, respectively; the actual displacement eccentricities 0, 0 are input into the two corresponding look-up table modules, respectively; reference displacements x, y in the X-axis and Y-axis directions are subtracted from the corresponding actual displacement eccentricities 0, 0 to obtain displacement errors ex, ey, respectively; initial control currents Ix0, Iy0 are obtained through the corresponding fuzzy adaptive PID controllers by using the displacement errors ex, ey, and actual control currents x, y are obtained through the corresponding amplifier modules by using the initial control currents Ix0, Iy0, respectively; the actual control currents x, y are input into the corresponding look-up table modules, and the two look-up table modules output the corresponding actual suspension forces x, y to the active magnetic bearing.



    [0008] Further, in Step (1), N levels of control currents and M levels of displacement eccentricities are selected for finite element simulation to obtain NM finite element models; the control currents, the displacement eccentricities, and the corresponding suspension forces of the NM finite element models in the X-axis and Y-axis directions are collected; the control currents and the displacement eccentricities in the X-axis and Y-axis directions of each of the finite element models are used as an independent variable x and the corresponding suspension forces are used as a dependent variable (x), and the independent variable x and the dependent variable (x) are substituted into a A general universal Kriging model y(x) = F(β, x) + z(x) to obtain two universal Kriging prediction models x(x, x̂0)=F1(β̂,îx, x̂0)+z1(îx, x̂0) and y(y,ŷ0)=F1(β̂,îy,ŷ0)+z1(y,ŷ0) through fitting, wherein F(β,x), F1(β̂,îx,0), and F1(β̂,îy,ŷ0) are regression models, z(x), z1(x, x̂0), and z1(y,ŷ0) are error terms, β is a regression coefficient of the general universal Kriging model, and β̂ is a regression coefficient of the universal Kriging prediction model.

    [0009] Further, the fuzzy adaptive PID controller in Step (3) consists of a fuzzy inference system, a proportional term, an integral term, and a derivative term; the displacement errors ex, ey and first-order derivatives x, y thereof are input into the corresponding fuzzy inference systems, and each of the fuzzy inference systems outputs a proportional modification coefficient CP, an integral modification coefficient CI, and a derivative modification coefficient CD; the proportional modification coefficient CP, the integral modification coefficient CI, and the derivative modification coefficient CD are respectively multiplied by a corresponding proportional coefficient KP, integral coefficient KI, and derivative coefficient KD to obtain a modified proportional term, integral term, and derivative term; a summation operation is performed on the outputs of the modified proportional term, integral term, and derivative term based on the displacement errors ex, ey to obtain the initial control currents Ix0, Iy0.

    [0010] The present invention has the following beneficial effects:
    1. 1. The present invention constructs accurately changing models of an active magnetic bearing under different displacement eccentricities and control currents based on the universal Kriging model theory and the suspension control principle of the active magnetic bearing, and obtains parameter tables for prediction of the actual suspension forces required by the active magnetic bearing along with the displacement change of the rotor. The parameter tables are created quickly, the tabulation cost is saved, more accurate models of the active magnetic bearing can be obtained according to actual situations, and the control accuracy is improved.
    2. 2. Compared with the conventional control of the active magnetic bearing, the present invention does not concern the fixed displacement stiffness and current stiffness in the process, so that its applicability is expanded from a pseudo-linear region near a working point to a non-linear region where both displacements and currents are large, and thus the control accuracy and the control range are improved.
    3. 3. Since the models are changeable, a common controller cannot make adjustments according to the models, and thus the present invention constructs a fuzzy adaptive PID control module, wherein errors and error rates are used as inputs on the basis of a PID algorithm, PID regulator parameters are changed through current control conditions, and fuzzy rules are used for fuzzy inference, so that the requirements on self-tuning of the PID parameters in accordance with errors and error rates at different times are met and more accurate control of the active magnetic bearing can be achieved.

    Brief Description of the Drawings



    [0011] 

    FIG. 1 is a block diagram of a fuzzy adaptive PID controller;

    FIG. 2 is a block diagram of a common PID controller; and

    FIG. 3 is a structural block diagram of an active magnetic bearing controller constructed by using a method of the present invention.


    Detailed Description of the Embodiments



    [0012] In the present invention, firstly, finite element models of an active magnetic bearing are built. Two universal Kriging prediction models in X-axis and Y-axis directions about actual suspension forces x, y being in association with actual displacement eccentricities 0, 0 and actual control currents x, y in the X-axis and Y-axis directions of the active magnetic bearing are obtained by using the finite element models of the active magnetic bearing. Two model state tables in the X-axis and Y-axis directions about the actual suspension forces x, y being in association with the actual displacement eccentricities 0, 0 and the actual control currents x, y are created. Based on the two model state tables, two look-up table modules with the two model state tables being built in are constructed, respectively. Two fuzzy adaptive PID controllers in the X-axis and Y-axis directions are constructed. Finally, the two fuzzy adaptive PID controllers, two corresponding amplifier modules in the X-axis and Y-axis directions, the two look-up table modules, and two corresponding measurement modules in the X-axis and Y-axis directions are used together to constitute an active magnetic bearing controller, which implements accurate control of the active magnetic bearing. The specific method is as follows:
    Dimension parameters of an active magnetic bearing to be controlled are measured, finite element models of the active magnetic bearing are built in finite element software, and performance parameters of the active magnetic bearing are obtained through simulation. On the premise of unsaturated magnetic field strength, N levels of control currents and M levels of displacement eccentricities are selected for finite element simulation to obtain NM finite element models, wherein N and M are selected according to the control currents, an air gap range, and the fineness of the models required. Then, data about the control currents in the X-axis and Y-axis directions, the displacement eccentricities in the X-axis and Y-axis directions, and the corresponding suspension forces in the X-axis and Y-axis directions of the NM finite element models are collected. The control currents and the displacement eccentricities in the X-axis and Y-axis directions of each model are a measured independent variable, and the corresponding suspension forces in the X-axis and Y-axis directions are a dependent variable. The present invention is described below by taking the X-axis direction as an example, and the implementation in the Y-axis direction is the same as that in the X-axis direction:
    Data about the control currents {i11, i12, ..., iNM}, the displacement eccentricities {x11, x12, ..., xNM}, and the corresponding suspension forces {F11, F12, ..., FNM} in the X-axis direction of the NM finite element models are collected. The control currents {i11, i12, ..., iNM} and the displacement eccentricities {x11, x12, ..., xNM} of each of the finite element models are a measured independent variable, and the suspension forces {F11, F12, ..., FNM} are a dependent variable. The independent variable can be expressed as Xij=[iij, xij]T, and the dependent variable can be expressed as Yij=Fij, wherein i=1, 2, ..., N and j=1, 2, ..., M.

    [0013] The general universal Kriging model is expressed as:

    wherein (x) is a final result value, that is, a dependent variable; F(β, x) is a regression model, wherein β is a regression coefficient and x is an independent variable of the universal Kriging model; z(x) is an error term in normal distribution with a mean of 0 and a variance of

    , wherein the variance

    is selected according to specific applications and will influence the accuracy of an approximate model. The regression model F(β, x) is expressed as:

    wherein β1, β2, ..., βp are regression coefficients of each order and fp(x) is a p-order approximate model.

    [0014] The independent variable Xij and the dependent variable Yij of the finite element models are used to respectively substitute x and (x) in the formula (1) of the general universal Kriging model. A universal Kriging prediction model in the X-axis direction about the actual suspension forces x being in association with the actual displacement eccentricities 0 and the actual control currents x in the X-axis direction of the active magnetic bearing can be obtained through fitting and is specifically expressed as:

    wherein β̂ is a regression coefficient of the built universal Kriging prediction model, and z1(x, x̂0) is an error term about the actual control currents x and the actual displacement eccentricities 0 in the X-axis direction.

    [0015] Similarly, the universal Kriging prediction model in the Y-axis direction is obtained as follows:

    wherein F1(β̂,y,0) is a regression model of the built universal Kriging prediction model, and z1(y, 0) is an error term about the actual control currents y and the actual displacement eccentricities 0 in the Y-axis direction.

    [0016] Two model state tables about the actual suspension forces x, y being in association with the actual displacement eccentricities 0, 0 and the actual control currents x, y are created according to the obtained two universal Kriging prediction models in the X-axis and Y-axis directions, respectively. Specifically, a model state table 1 about the actual suspension forces x being in association with the actual displacement eccentricities 0 and the actual control currents x in the X-axis direction is built, and a model state table 2 about the actual suspension forces y being in association with the actual displacement eccentricities 0 and the actual control currents y in the Y-axis direction is built:
    Table 1
    Actual displacement eccentricities/ control currents in the X-axis direction00.010.020.03···xmax
    0 F11 F12 F13 F14 ··· F1a
    0.1 F21 F22 F23 F24 ··· F2a
    0.2 F31 F32 F33 F34 ··· F3a
    0.3 F41 F42 F43 F44 ··· F4a
    imax Fb1 Fb2 Fb3 Fb4 ··· Fba
    Table 2
    Actual displacement eccentricities/ control currents in the Y-axis direction00.010.020.03···ymax
    0 F'11 F'12 F'13 F'14 ··· F'1a
    0.1 F'21 F'22 F'23 F'24 ··· F'2a
    0.2 F'31 F'32 F'33 F'34 ··· F'3a
    0.3 F'41 F'42 F'43 F'44 ··· F'4a
    iymax F'b1 F'b2 F'b3 F'b4 ··· F'ba


    [0017] Taking Table 1 as an example, the first row is about the actual displacement eccentricities 0 in the X-axis direction, and the first column is about the actual control currents x. The actual displacement eccentricities 0 start from 0 to the maximum eccentricity xmax and are sampled at an interval of 0.01 mm; meanwhile, the actual control currents x start from 0 and are sampled at an interval of 0.1 A. The sampled values of the actual displacement eccentricities 0 and the actual control currents x are substituted into the formula (3) to calculate the actual suspension forces x associated with the sampled values, and the model state table 1 is created accordingly. Therefore, each of the actual displacement eccentricities 0 and each of the actual control currents x are corresponding to one of the actual suspension forces x , as shown by F11 to Fba in Table 1, wherein xmax is a maximum displacement in the X-axis direction and imax is a maximum control current. Taking Table 1 as an example, when the actual displacement eccentricity is 0.01 mm and the actual control current is 0.1 A, the actual suspension force is F22; and when the actual displacement eccentricity is 0.03 mm and the actual control current is 0.2 A, the actual suspension force is F34. In Table 1, b and a are respectively the number of the sampled actual displacement eccentricities and the number of the sampled actual control currents. Similarly, in Table 2, the actual suspension forces are F'11 to F'ba, ymax is a maximum displacement in the Y-axis direction, and iymax is a maximum control current in the Y-axis direction. Similarly, in Table 2, the first row is about the actual displacement eccentricities 0 in the Y-axis direction, and the first column is about the actual control currents y. The sampling mode is the same as that in the X-axis direction. The sampled values of the actual displacement eccentricities 0 and the actual control currents y are substituted into the formula (3) to calculate the actual suspension forces y associated with the sampled values, and the model state table 2 is created accordingly.

    [0018] As for data of the actual displacement eccentricities and the actual control currents that are not sampled values, the corresponding actual suspension forces are calculated through interpolation. Taking the X-axis direction as an example, when the actual displacement eccentricity is x0 and the actual control current is i0, the positions of x0 and i0 in Table 1 need to be determined first. Assume that the displacement eccentricity of {x0, i0} falls between the sampled values x1 and x2, the control current falls between the sampled values i1 and i2, x1 and x2 satisfy x1+0.01mm=x2, i1 and i2 satisfy i1+0.1A=i2, and x1, x2, i1, and i2 are all sampled displacements and currents. At this time, the actual suspension force corresponding to the sampled values {x1, i1} is Fc,d, the actual suspension force corresponding to {x1, i2} is Fc,d+1, the actual suspension force corresponding to {x2, i1} is Fc+1,d, and the actual suspension force corresponding to {x2, i2} is Fc+1,d+1, wherein c and d are the row number and the column number of the sampled values {x1, i1} in Table 1. Then, the actual suspension force corresponding to {x0, i0} can be calculated as:



    [0019] For example, when the actual displacement eccentricity is 0.025 mm and the actual control current is 0.25 A, the corresponding suspension force can be calculated according to data in Table 1 as follows:



    [0020] Similarly, as for data of the actual displacement eccentricities and the actual control currents in the Y-axis direction that are not sampled values, the corresponding actual suspension forces are calculated through interpolation in the same way.

    [0021] Two look-up table modules are constructed, wherein the model state table 1 and the model state Table 2 are respectively built in the look-up table module in the X-axis direction and the look-up table module in the Y-axis direction.

    [0022] A fuzzy adaptive PID controller shown in FIG. 1 is constructed. Since changes of input currents and displacements of the active magnetic bearing may cause certain errors in the models, a common controller cannot make adjustments according to the models, and thus the present invention adopts a fuzzy adaptive PID controller for control. FIG. 2 is a structural block diagram of an existing common PID controller, which mainly consists of a proportional term, an integral term, and a derivative term. The proportional term is directly composed of a proportional coefficient KP, the integral term is directly composed of an integral coefficient KI and an integral module ∫, and the derivative term is directly composed of a derivative coefficient KD and a derivative module d/dt. A summation operation, denoted by ∑, is performed on the three terms to obtain a final output. FIG. 1 shows the fuzzy adaptive PID controller in the X-axis direction that is constructed by the present invention and consists of a fuzzy inference system, a proportional term, an integral term, and a derivative term. Compared with the existing common PID controller shown in FIG. 2, improvements have been made in the controller in FIG. 1, and the fuzzy inference system is employed in addition to the proportional term, the integral term, and the derivative term in the PID controller in FIG. 2. Taking the fuzzy adaptive PID controller in the X-axis direction as an example, displacement errors ex in the X-axis direction and first-order derivatives x thereof are input into the fuzzy inference system, and the fuzzy inference system calculates according to the displacement errors ex and the first-order derivatives x to output a proportional modification coefficient CP, an integral modification coefficient CI, and a derivative modification coefficient CD. The proportional modification coefficient CP, the integral modification coefficient CI, and the derivative modification coefficient CD are respectively multiplied by the corresponding proportional coefficient KP, integral coefficient KI, and derivative coefficient KD to obtain the modified proportional coefficient, the modified integral coefficient, and the modified derivative coefficient, which are expressed as:



    [0023] A summation operation, denoted by ∑, is performed on the modified proportional term, integral term, and derivative term to obtain a final output. That is, initial control currents Ix0 output in the X-axis direction can be accurately controlled through the modified proportional term, integral term, and derivative term based on the displacement errors ex.

    [0024] According to the influence of parameter adjustment on the output performance of the system, the modification coefficients CP, CI, CD are adjusted based on the following principles: When ex is large,

    is increased,

    is decreased, and

    is kept moderate through the modification coefficients to improve the response speed of the system and meanwhile prevent excessive overshoot. When ex is moderate,

    and

    are kept small while

    is kept moderate through the modification coefficients to reduce the overshoot and meanwhile enable the system to respond quickly. When ex is small,

    and

    are increased while

    is kept moderate through the modification coefficients to ensure good stability of the system, avoid system oscillation, and enhance the anti-interference performance of the system.

    [0025] Similarly, the fuzzy adaptive PID controller in the Y-axis direction is constructed in the same way as the fuzzy adaptive PID controller in the X-axis direction. Displacement errors ey in the Y-axis direction and first-order derivatives y thereof are input into the corresponding fuzzy inference system in the Y-axis direction, and the fuzzy inference system outputs a proportional modification coefficient CP, an integral modification coefficient CI, and a derivative modification coefficient CD. The proportional modification coefficient CP, the integral modification coefficient CI, and the derivative modification coefficient CD are respectively multiplied by the corresponding proportional coefficient KP, integral coefficient KI, and derivative coefficient KD to obtain the modified proportional term, integral term, and derivative term. A summation operation is performed on the outputs of the modified proportional term, integral term, and derivative term based on the displacement errors ey to obtain initial control currents Iy0 in the Y-axis direction.

    [0026] An active magnetic bearing controller shown in FIG. 3 is constructed. The active magnetic bearing controller consists of two fuzzy adaptive PID controllers, two amplifier modules, two look-up table modules, and two measurement modules in the X-axis and Y-axis directions, and is connected to an input end of the active magnetic bearing to implement control of the active magnetic bearing. The fuzzy adaptive PID controller, the amplifier module, and the look-up table module in the X-axis direction are connected in series and then connected to the input end of the active magnetic bearing. The fuzzy adaptive PID controller, the amplifier module, and the look-up table module in the Y-axis direction are connected in series and then connected to the input end of the active magnetic bearing. The two measurement modules in the X-axis and Y-axis directions measure, through displacement sensors, the actual displacement eccentricities 0, 0 in the X-axis and Y-axis directions of the active magnetic bearing, respectively. The actual displacement eccentricities 0 in the X-axis direction are input into the look-up table module in the X-axis direction, and the actual displacement eccentricities 0 in the Y-axis direction are input into the look-up table module in the Y-axis direction. The reference displacements x are subtracted from the actual displacement eccentricities 0 to obtain the displacement errors ex in the X-axis direction. The initial control currents Ix0 are obtained through the fuzzy adaptive PID controller in the X-axis direction by using the displacement errors ex, and the actual control currents x are obtained through the amplifier module in the X-axis direction. The actual control currents x are input into the look-up table module in the X-axis direction, and the look-up table module in the X-axis direction obtains the actual suspension forces x according to data in the model state table 1. Similarly, the reference displacements y are subtracted from the actual displacement eccentricities 0 to obtain the displacement errors ey in the Y-axis direction. The initial control currents Iy0 are obtained through the fuzzy adaptive PID controller in the Y-axis direction by using the displacement errors ey, and the actual control currents y are obtained through the amplifier module in the Y-axis direction. The actual control currents y are input into the look-up table module in the Y-axis direction, and the look-up table module in the Y-axis direction obtains the actual suspension forces y according to data in the model state table 2 and outputs the actual suspension forces y to the active magnetic bearing. That is, the look-up table modules in the X-axis and Y-axis directions respectively output the corresponding actual suspension forces x, y to the active magnetic bearing, to implement control of the active magnetic bearing in the X-axis and Y-axis directions.


    Claims

    1. A method for constructing an active magnetic bearing controller based on a look-up table method, characterized by comprising the following steps:

    Step (1): building finite element models of an active magnetic bearing, and obtaining, by using the finite element models and based on a general universal Kriging model, two universal Kriging prediction models about actual suspension forces x, y being in association with actual displacement eccentricities 0, 0 and actual control currents x, y in X-axis and Y-axis directions;

    Step (2): creating, based on the two universal Kriging prediction models, two model state tables about the actual suspension forces x, y being in association with the actual displacement eccentricities 0, 0 and the actual control currents x, y to construct two corresponding look-up table modules with the two model state tables being built in, respectively; and

    Step (3): constructing an active magnetic bearing controller by using two fuzzy adaptive PID controllers, two amplifier modules, two look-up table modules, and two measurement modules in the X-axis and Y-axis directions, wherein the fuzzy adaptive PID controller, the amplifier module, and the look-up table module in the X-axis direction are connected in series and then connected to an input end of the active magnetic bearing; the fuzzy adaptive PID controller, the amplifier module, and the look-up table module in the Y-axis direction are connected in series and then connected to the input end of the active magnetic bearing; the two measurement modules in the X-axis and Y-axis directions measure the actual displacement eccentricities 0, 0 of the active magnetic bearing in the X-axis and Y-axis directions, respectively; the actual displacement eccentricities 0, 0 are input into the two corresponding look-up table modules, respectively; reference displacements x, y in the X-axis and Y-axis directions are subtracted from the corresponding actual displacement eccentricities 0, 0 to obtain displacement errors ex, ey, respectively; initial control currents Ix0, Iy0 are obtained through the corresponding fuzzy adaptive PID controllers by using the displacement errors ex, ey, and actual control currents x, y are obtained through the corresponding amplifier modules by using the initial control currents Ix0, Iy0, respectively; the actual control currents x, y are input into the corresponding look-up table modules, and the two look-up table modules output the corresponding actual suspension forces x, y to the active magnetic bearing.


     
    2. The method for constructing the active magnetic bearing controller based on the look-up table method according to claim 1, characterized in that in Step (1), N levels of control currents and M levels of displacement eccentricities are selected for finite element simulation to obtain NM finite element models; the control currents, the displacement eccentricities, and the corresponding suspension forces of the NM finite element models in the X-axis and Y-axis directions are collected; the control currents and the displacement eccentricities in the X-axis and Y-axis directions of each of the finite element models are used as an independent variable x and the corresponding suspension forces are used as a dependent variable (x), and the independent variable x and the dependent variable (x) are substituted into a general universal Kriging model (x) = F(β, x) + z(x) to obtain two universal Kriging prediction models x(x, 0)=F1(β̂,x, 0)+z1(x, 0) and y(y,ŷ0)=F1(β̂,îy,0)+z1(y,ŷ0) through fitting, wherein F(β,x), F1(β̂,îx,x̂0), and F1(β̂,îy,ŷ0) are regression models, z(x), z1(x, 0), and z1(y, 0) are error terms, β is a regression coefficient of the general universal Kriging model, and β̂ is a regression coefficient of the universal Kriging prediction model.
     
    3. The method for constructing the active magnetic bearing controller based on the look-up table method according to claim 2, characterized in that the actual displacement eccentricities 0, 0 start from 0 and are sampled at an interval of 0.01 mm, the actual control currents x, y start from 0 and are sampled at an interval of 0.1 A, the sampled values of the actual displacement eccentricities 0, 0 and the actual control currents x, y are substituted into the corresponding universal Kriging prediction models to calculate the actual suspension forces x, y, to create the corresponding two model state tables.
     
    4. The method for constructing the active magnetic bearing controller based on the look-up table method according to claim 3, characterized in that first rows of the two model state tables are about the actual displacement eccentricities 0, 0 and first columns of the two model state tables are about the actual control currents x, y, respectively; and each of the actual displacement eccentricities 0, 0 and each of the actual control currents x, y are corresponding to one of the actual suspension forces x, y.
     
    5. The method for constructing the active magnetic bearing controller based on the look-up table method according to claim 3, characterized in that as for data of the actual displacement eccentricities and the actual control currents that are not sampled values, the corresponding actual suspension forces are calculated through interpolation.
     
    6. The method for constructing the active magnetic bearing controller based on the look-up table method according to claim 1, characterized in that the fuzzy adaptive PID controller in Step (3) consists of a fuzzy inference system, a proportional term, an integral term, and a derivative term; the displacement errors ex, ey and first-order derivatives x, y thereof are input into the corresponding fuzzy inference systems, and each of the fuzzy inference systems outputs a proportional modification coefficient CP, an integral modification coefficient CI, and a derivative modification coefficient CD; the proportional modification coefficient CP, the integral modification coefficient CI, and the derivative modification coefficient CD are respectively multiplied by a corresponding proportional coefficient KP, integral coefficient KI, and derivative coefficient KD to obtain a modified proportional term, integral term, and derivative term; a summation operation is performed on the outputs of the modified proportional term, integral term, and derivative term based on the displacement errors ex, ey to obtain the initial control currents Ix0, Iy0.
     


    Ansprüche

    1. Verfahren zum Konstruieren eines aktiven Magnetlagerreglers auf der Grundlage eines Nachschlagetabellenverfahrens, dadurch gekennzeichnet, dass es die folgenden Schritte umfasst:

    Schritt (1): Erstellen von Finite-Elemente-Modellen eines aktiven Magnetlagers und Erhalten, unter Verwendung der Finite-Elemente-Modelle und auf der Grundlage eines allgemeinen universellen Kriging-Modells, von zwei universellen Kriging-Vorhersagemodellen über die tatsächlichen Aufhängungskräfte x, y in Verbindung mit tatsächlichen Verschiebungsexzentrizitäten 0, 0 und tatsächliche Steuerströme x, y in X-Achsen- und Y-Achsen-Richtungen;

    Schritt (2): Erstellen, auf der Grundlage der beiden universellen Kriging-Vorhersagemodellen, von zwei Modellzustandstabellen über die tatsächlichen Aufhängungskräfte x, y in Verbindung mit den tatsächlichen Verschiebungsexzentrizitäten 0, 0 und den tatsächlichen Steuerströmen x, y, um jeweils zwei entsprechende Nachschlagetabellenmodule mit den beiden eingebauten Modellzustandstabellen zu konstruieren,

    Schritt (3): Konstruieren eines aktiven Magnetlagerreglers unter Verwendung von zwei adaptiven Fuzzy-PID-Reglern, zwei Verstärkermodulen, zwei Nachschlagetabellenmodulen und zwei Messmodulen in der X-Achsen- und Y-Achsen-Richtung, wobei der adaptive Fuzzy-PID-Regler, das Verstärkermodul und das Nachschlagetabellenmodul in der X-Achsen-Richtung in Reihe geschaltet und dann mit einem Eingangsende des aktiven Magnetlagers verbunden werden; wobei der adaptive Fuzzy-PID-Regler, das Verstärkermodul und das Nachschlagetabellenmodul in der Y-Achsen-Richtung in Reihe geschaltet und dann mit dem Eingangsende des aktiven Magnetlagers verbunden sind; wobei die beiden Messmodule in X-Achsen- und Y-Achsen-Richtung jeweils die tatsächlichen Verschiebungsexzentrizitäten 0, 0 des aktiven Magnetlagers in X-Achsen- bzw. Y-Achsen-Richtung messen; wobei die tatsächlichen Verschiebungsexzentrizitäten 0, 0 jeweils in die beiden entsprechenden Nachschlagetabellenmodule eingegeben werden; wobei Referenzverschiebung x, y in X-Achsen- und Y-Achsen-Richtung jeweils von den entsprechenden tatsächlichen Verschiebungsexzentrizitäten 0, 0 subtrahiert werden, um Verschiebungsfehler ex, ey zu erhalten; wobei die anfänglichen Steuerströme Ix0, Iy0 durch die entsprechenden adaptiven Fuzzy-PID-Regler unter Verwendung der Verschiebungsfehler ex, ey gewonnen werden, wobei die tatsächliche Steuerströme x, y jeweils durch die entsprechenden Verstärkermodule unter Verwendung der anfänglichen Steuerströme Ix0, Iy0 gewonnen werden; wobei die tatsächlichen Steuerströme x, y in die entsprechenden Nachschlagetabellenmodul eingegeben werden, und wobei die beiden Nachschlagetabellenmodul die entsprechenden tatsächlichen Aufhängungskräfte x, y an das aktive Magnetlager ausgeben.


     
    2. Verfahren zum Konstruieren eines aktiven Magnetlagerreglers auf der Grundlage eines Nachschlagetabellenverfahrens nach Anspruch 1,dadurch gekennzeichnet, dass in Schritt (1) N Niveaus von Steuerströmen und M Niveaus von Verschiebungsexzentrizitäten für die Finite-Elemente-Simulation ausgewählt werden, um NM Finite-Elemente-Modelle zu erhalten; wobei die Steuerströme, die Verschiebungsexzentrizitäten und die entsprechenden Aufhängungskräfte der NM Finite-Elemente-Modelle in den X-Achsen- und Y-Achsen-Richtungen gesammelt werden; wobei die Steuerströme und die Verschiebungsexzentrizitäten in den X-Achsen- und Y-Achsen-Richtungen jedes der Finite-Elemente-Modellen als unabhängige Variable x und die entsprechenden Aufhängungskräfte als abhängige Variable (x) verwendet werden, und wobei die unabhängige Variable x und die abhängige Variable (x) in ein allgemeines universelles Kriging-Modell (x) = F(β, x) + z(x) eingesetzt werden, um zwei universelle Kriging-Vorhersagemodelle x(x, 0)=F1(β̂,x, 0)+z1(x, 0) und y(y, 0)=F1(β̂,y,0)+z1(y, 0) durch Anpassung zu erhalten, wobei F(β,x), F1(β̂,x,0) und F1(β̂,y,0) Regressionsmodelle sind, wobei z(x), z1(x, 0), und z1(iy, 0) Fehlerterme sind, und wobei β ein Regressionskoeffizient des allgemeinen universellen Kriging-Modells ist und β̂ ein Regressionskoeffizient des universellen Kriging-Vorhersagemodells ist.
     
    3. Verfahren zum Konstruieren eines aktiven Magnetlagerreglers auf der Grundlage eines Nachschlagetabellenverfahrens nach Anspruch 2, dadurch gekennzeichnet, dass die tatsächlichen Verschiebungsexzentrizitäten 0, 0 bei 0 beginnen und in einem Intervall von 0,01 mm abgetastet werden, wobei die tatsächliche Steuerströme x, y bei 0 beginnen und in einem Intervall von 0,1 A abgetastet werden, wobei die abgetasteten Werte der tatsächlichen Verschiebungsexzentrizitäten 0, 0 und der tatsächlichen Steuerströme x, y in die entsprechenden universellen Kriging-Vorhersagemodelle eingesetzt werden, um die tatsächlichen Aufhängungskräfte x, y zu berechnen, um die entsprechenden zwei Modellzustandstabellen zu erzeugen.
     
    4. Verfahren zum Konstruieren eines aktiven Magnetlagerreglers auf der Grundlage eines Nachschlagetabellenverfahrens nach Anspruch 3, dadurch gekennzeichnet, dass die erste Zeilen der beiden Modellzustandstabellen jeweils die tatsächlichen Verschiebungsexzentrizitäten 0, 0 und die erste Spalten der beiden Modellzustandstabellen die tatsächlichen Steuerströme x, îy betreffen; und jede der tatsächlichen Verschiebungsexzentrizitäten 0, 0 und jeder der tatsächlichen Steuerströme x, y einer der tatsächlichen Aufhängungskräfte x, y entspricht.
     
    5. Verfahren zum Konstruieren eines aktiven Magnetlagerreglers auf der Grundlage eines Nachschlagetabellenverfahrens nach Anspruch 3, dadurch gekennzeichnet, dass für Daten der tatsächlichen Verschiebungsexzentrizitäten und der tatsächlichen Steuerströme, die keine Abtastwerte sind, die entsprechenden tatsächlichen Aufhängungskräfte durch Interpolation berechnet werden.
     
    6. Verfahren zum Konstruieren eines aktiven Magnetlagerreglers auf der Grundlage eines Nachschlagetabellenverfahrens nach Anspruch 1, dadurch gekennzeichnet, dass der adaptive Fuzzy-PID-Regler in Schritt (3) aus einem Fuzzy-Inferenzsystem, einem Proportionalterm, einem Integralterm und einem Ableitungsterm besteht; wobei die Verschiebungsfehler ex, ey und deren Ableitungen x, y erster Ordnung in die entsprechenden Fuzzy-Inferenzsysteme eingegeben werden, wobei jedes der Fuzzy-Inferenzsysteme einen Proportionalmodifikationskoeffizienten CP, einen Integralmodifikationskoeffizienten CI und einen Ableitungsmodifikationskoeffizienten CD ausgibt; wobei der Proportionalmodifikationskoeffizient CP, der Integralmodifikationskoeffizient CI und der Ableitungsmodifikationskoeffizient CD jeweils mit einem entsprechenden Proportionalkoeffizienten KP, Integralkoeffizienten KI und Ableitungskoeffizienten KD multipliziert werden, um einen modifizierten Proportionalterm, Integralterm und Ableitungsterm zu erhalten; wobei eine Summationsoperation an den Ausgängen des modifizierten Proportionalterms, Integralterms und Ableitungsterms auf der Grundlage der Verschiebungsfehler ex, ey durchgeführt wird, um die anfänglichen Steuerströme Ix0, Iy0 zu erhalten.
     


    Revendications

    1. Procédé pour construire un contrôleur de palier magnétique actif basé sur un procédé de table de consultation, caractérisé en ce qu'il comprend les étapes suivantes :

    étape (1) : établir des modèles d'éléments finis d'un palier magnétique actif, et obtenir, à travers les modèles d'éléments finis et basé sur un modèle universel général de Krigeage, deux modèles universels de prédiction de Krigeage sur forces de suspension réelles x et y en association avec des excentricités de déplacement réelles 0 et 0, et des courants de commande réels x et y dans des directions d'axe X et d'axe Y ;

    étape (2) : établir, basé sur les deux modèles universels de prédiction de krigeage, deux tables d'états de modèle sur les forces de suspension réelles x et y en association avec les excentricités de déplacement réelles 0 et 0, et les courants de commande réels x et y, de sorte d'établir respectivement deux modules de table de consultation correspondants aux deux tables d'états de modèle intégrées ; et

    étape (3) : construire un contrôleur de palier magnétique actif en utilisant deux contrôleurs PID adaptatifs flous, deux modules d'amplificateur, deux modules de table de consultation et deux modules de mesure dans les directions d'axe X et d'axe Y, le contrôleur PID adaptatif flou, le module d'amplificateur et le module de table de consultation dans les directions d'axe X étant connectés en série et puis connectés à une extrémité d'entrée du palier magnétique actif ; le contrôleur PID adaptatif flou, le module amplificateur et le module de table de consultation dans la direction d'axe Y étant connectés en série et puis connectés à l'extrémité d'entrée du palier magnétique actif ; les deux modules de mesure dans les directions d'axe X et d'axe Y mesurant respectivement les excentricités de déplacement réelles 0 et 0 du palier magnétique actif dans les directions d'axe X et d'axe Y ; les excentricités de déplacement réelles 0 et 0 étant entrées respectivement dans les deux modules de table de consultation correspondants ; les déplacements de référence x et y dans les directions d'axe X et d'axe Y étant soustraits des excentricités de déplacement réelles correspondantes 0 et 0 de sorte d'obtenir respectivement les erreurs de déplacement ex et ey ; les courants de commande initiaux Ix0 et Iy0 étant obtenus par les contrôleurs PID adaptatifs flous correspondants en utilisant les erreurs de déplacement ex et ey, et les courants de commande réels x et y étant obtenus par les modules d'amplificateur correspondants en utilisant respectivement les courants de commande initiaux Ix0 et Iy0; les courants de commande réels x et y étant entrés dans les modules de table de consultation correspondants, et les deux modules de table de consultation délivrant les forces de suspension réelles correspondantes x et y au palier magnétique actif.


     
    2. Procédé pour construire le contrôleur de palier magnétique actif basé sur le procédé de table de consultation selon la revendication 1, caractérisé en ce que, à l'étape (1), N niveaux de courants de commande et M niveaux d'excentricités de déplacement sont sélectionnés pour simuler des éléments finis, de sorte d'obtenir NM modèles d'éléments finis ; les courants de commande, les excentricités de déplacement et les forces de suspension correspondantes des NM modèles d'éléments finis dans les directions d'axes X et d'axe Y sont collectés ; les courants de commande et les excentricités de déplacement dans les directions d'axes X et d'axe Y de chacun des modèles d'éléments finis sont utilisés comme variable indépendante x et les forces de suspension correspondantes sont utilisées comme variable dépendante (x), et la variable indépendante x et le variables dépendantes (x) sont substituées dans un modèle universel général de Krigeage (x) = F(β, x) + z(x) de sorte d'obtenir deux modèles universels de prédiction de Krigeage x(x, 0)=F1(β̂,îx, 0)+z1(x, 0) et y(y,0)=F1(β̂,îy,0)+z1(y,0) en ajustant, où F(β,x), F1(β̂,îx,0) et F1(β̂,y,0) sont des modèles de régression, z(x), z1(îx, x̂0) et z1(y, 0) sont des termes d'erreur, β est un coefficient de régression du modèle universel général de Krigeage, et β̂ est un coefficient de régression du modèle universel général de Krigeage.
     
    3. Procédé pour construire le contrôleur de palier magnétique actif basé sur le procédé de table de correspondance selon la revendication 2, caractérisé en ce que les excentricités de déplacement réelles 0 et 0 partent de 0 et sont échantillonnées à un intervalle de 0,01 mm, les courants de commande réels x et y partent de 0 et sont échantillonnés à un intervalle de 0,1 A, les valeurs échantillonnées des excentricités de déplacement réelles 0 et 0, et les courants de commande réels x et y sont substitués dans les modèles universels de prédiction de krigeage correspondants pour calculer les forces de suspension réelles x et y, de sorte d'établir les deux tables d'états de modèle correspondants.
     
    4. Procédé pour construire le contrôleur de palier magnétique actif basé sur le procédé de table de consultation selon la revendication 3, caractérisé en ce que les premières rangs des deux tables d'états de modèle correspondent respectivement aux excentricités de déplacement réelles 0 et 0, et les premières colonnes des deux tables d'états de modèle correspondent respectivement aux courants de commande réels x et y ; et chacune des excentricités de déplacement réelles 0 et 0 et chacun des courants de commande réels x et y correspondent à l'une des forces de suspension réelles x et y.
     
    5. Procédé pour construire le contrôleur de palier magnétique actif basé sur le procédé de table de consultation selon la revendication 3, caractérisé en ce que, pour les données des excentricités de déplacement réelles et des courants de commande réels qui ne sont pas des valeurs échantillonnées, les forces de suspension réelles correspondantes sont calculé par interpolation.
     
    6. Procédé pour construire le contrôleur de palier magnétique actif basé sur le procédé de table de consultation selon la revendication 1, caractérisé en ce que le contrôleur PID adaptatif flou à l'étape (3) comprend un système d'inférence flou, un terme proportionnel, un terme intégrale, et un terme dérivé ; les erreurs de déplacement ex, ey et leurs dérivées de premier ordre x et y sont entrées dans les systèmes d'inférence floue correspondants, et chacun des systèmes d'inférence floue délivre un coefficient de modification proportionnel CP, un coefficient de modification intégrale CI et un coefficient de modification dérivée CD ; le coefficient de modification proportionnel CP, le coefficient de modification intégrale CI et le coefficient de modification dérivée CD sont respectivement multipliés par un coefficient proportionnel correspondant KP, un coefficient intégral KI et un coefficient dérivé KD pour obtenir un terme proportionnel modifié, un terme intégral et un terme dérivé ; une opération de sommation est effectuée sur les sorties du terme proportionnel modifié, du terme intégral et du terme dérivé basé sur les erreurs de déplacement ex et ey pour obtenir les courants de commande initiaux Ix0 et Iy0.
     




    Drawing














    Cited references

    REFERENCES CITED IN THE DESCRIPTION



    This list of references cited by the applicant is for the reader's convenience only. It does not form part of the European patent document. Even though great care has been taken in compiling the references, errors or omissions cannot be excluded and the EPO disclaims all liability in this regard.

    Patent documents cited in the description