JP2001165163A - Magnetic bearing control device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a magnetic bearing control device capable of easily and stably controlling a rotating shaft and a magnetic bearing and easily designed and manufactured. SOLUTION: This magnetic baring control device is provided with a difference means 7 for obtaining deviation δ1 between the displacement (x) of a rotating shaft detected by a sensor 4, and the target value (r) of displacement; a differential means 75 for obtaining the differential value δ2 of the deviation; a first gain means 74 for selecting gains K1a, K1b determined on the basis of the predetermined fuzzy model of a rotating shaft.magnetic bearing system, according to the deviation δ1 and the differential value δ2 of the deviation and multiplying the deviation δ1 by the selected gain; a second gain means 76 for selecting gains K2a, K2b determined on the basis of the fuzzy model of the rotating shaft.magnetic bearing system, according to the deviation δ1 and the differential value δ2 of the deviation and multiplying the differential value δ2 of the deviation by the selected gain; and an adding means 7 for adding the value computed by the first gain means 74 and the value computed by the second gain means 76 and outputting the added value.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a magnetic bearing for supporting a rotating shaft by a magnetic force of an electromagnet, and more particularly, to controlling an exciting voltage or an exciting current of the electromagnet so as to support the rotating shaft at a target position. The present invention relates to a magnetic bearing control device.

[0002]

2. Description of the Related Art A magnetic bearing is characterized in that the rotating shaft can be supported in a completely non-contact state by a magnetic force, does not wear the bearing, and does not require lubricating oil or the like. Due to such characteristics, the magnetic bearing is often used as a bearing for a device in which a rotating shaft rotates at a high speed or a device used in a special atmosphere (for example, under a condition of high temperature, low temperature, vacuum, or the like).

In such a magnetic bearing, it is necessary to keep the rotating shaft levitated by magnetic force at all times. Therefore, the magnetic bearing controls the exciting voltage or exciting current of the electromagnet so as to support the rotating shaft at a target position. A magnetic bearing control device is provided. The magnetic bearing control device includes an electromagnet that floats and supports the rotating shaft by exerting a magnetic attraction force on the rotating shaft, a sensor that detects a floating position of the rotating shaft, and a target for the rotating shaft based on a signal from the sensor. And a controller for controlling the exciting voltage or exciting current (ie, magnetic attractive force) of the electromagnet so as to be supported at the position.

Conventionally, controllers such as a PID (proportional-integral-derivative) controller and a phase compensator using an analog control method have been widely used as controllers of the magnetic bearing control device.

[0005]

However, non-linearity when the rotating shaft is greatly displaced, displacement detection circuit, control compensation circuit, power amplifier, frequency characteristics of the coil in the electromagnet, free vibration of the rotating shaft, and imbalance. Due to various causes such as forced vibration and gyro effect, the dynamic characteristics of the rotating shaft / magnetic bearing system are very complicated. Therefore, when the magnetic bearing is controlled by analog control as in the related art, it takes considerable time and effort to set and adjust the control parameters.

In particular, since the analog controller is realized by a combination of hardware such as an operational amplifier, a resistor, and a capacitor, when the analog controller is used as a controller of the magnetic bearing control device, the rotation shaft and the magnetic field are not controlled. When it becomes necessary to modify or adjust the control system of the bearing, it is necessary to change these hardware. Therefore,
It takes a considerable amount of time and effort to correct and adjust the control system for the rotating shaft and magnetic bearing.

When a controller for PID control or only phase compensation is used as a controller of the magnetic bearing control device,
Since the robustness of the control system is low, the control system is likely to become unstable when the dynamic characteristics of the rotating shaft / magnetic bearing system change or when disturbance is mixed in the control system.

Further, in designing and manufacturing the magnetic bearing control device, it is necessary to perform modeling (identification) of the rotating shaft and the magnetic bearing to be controlled, and the dynamic characteristics of the rotating shaft and the magnetic bearing system are as described above. And it is difficult to determine an exact mathematical model of the control system.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems of the prior art, and it is possible to easily and stably control a rotating shaft and a magnetic bearing, and to design and manufacture the same easily. It is an object of the present invention to provide a magnetic bearing control device capable of performing the following.

[0010]

In order to solve the problems in the prior art, the present invention according to claim 1 sets a reference point in a magnetic bearing that supports a rotating shaft by a magnetic force of an electromagnet. A magnetic bearing control device comprising a sensor for detecting the displacement of the rotating shaft from the reference point and controlling a current or a voltage supplied to the electromagnet based on the displacement of the rotating shaft detected by the sensor. ,
Difference means for calculating a deviation between the displacement of the rotation axis detected by the sensor and a target value of the displacement of the rotation axis, and differentiation means for calculating a differential value of the deviation obtained by the difference means;
A gain determined based on a predetermined fuzzy model of the rotating shaft / magnetic bearing system is selected in accordance with the deviation obtained by the difference means and the differential value of the deviation obtained by the differentiation means. A first gain means for multiplying the deviation, a gain determined based on a predetermined fuzzy model of the rotating shaft / magnetic bearing system, a deviation obtained by the difference means and a deviation obtained by the differentiation means. Second gain means for selecting according to the differential value of the difference, multiplying by the differential value of the deviation, and adding the value calculated by the first gain means and the value calculated by the second gain means. And an adder that outputs the value as a current or voltage to be supplied to the electromagnet. Thus, the rotating shaft / magnetic bearing system can be controlled by the sliding mode control.

According to a second aspect of the present invention, in the first aspect of the present invention, the displacement of the rotating shaft, the differential value of the displacement of the rotating shaft, and the current or voltage supplied to the electromagnet are determined. A fuzzy neural network is configured for the fuzzy model of the rotating shaft / magnetic bearing system described including, and the fuzzy neural network supplies the displacement of the rotating shaft, the differential value of the displacement of the rotating shaft, and the supply to the electromagnet. By inputting the current or voltage to be applied, the displacement of the rotating shaft after the elapse of the unit time is estimated, and the estimated displacement of the rotating shaft after the elapse of the unit time and the actual displacement of the rotating shaft after the elapse of the unit time are input. The difference is obtained, the fuzzy neural network is repeatedly learned so as to minimize the difference, and an accurate fuzzy model of the rotating shaft / magnetic bearing system is obtained based on the learned result. Came out, the rotation axis-derived said
The gain is determined for a fuzzy model of a magnetic bearing system. As a result, an accurate fuzzy model of the rotating shaft / magnetic bearing system can be obtained by the fuzzy neural network, and a sliding mode control based on the accurate fuzzy model of the rotating shaft / magnetic bearing system can be performed.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an embodiment of a magnetic bearing control device according to the present invention will be described in detail with reference to the drawings. FIG. 1 is a diagram illustrating a configuration of a magnetic bearing according to the present embodiment. As shown in FIG.
Is provided, and the rotating shaft 1 can be rotated at a high speed. Radial magnetic bearings 2 are arranged at two positions in the axial direction of the rotating shaft 1, and thrust magnetic bearings are arranged near both ends of a disk portion 12 attached to one end of the rotating shaft 1.
The radial magnetic bearing 2 supports the rotating shaft 1 in a radial direction by a magnetic force, and the thrust magnetic bearing 3 supports the rotating shaft 1 in an axial direction by a magnetic force. The radial magnetic bearing 2 and the thrust magnetic bearing 3 form a five-axis control magnetic bearing.

Each of the bearings 2 and 3 is an electromagnet which floats and supports the rotating shaft 1 by applying a magnetic attraction to the rotating shaft 1.
A gap sensor 4 for detecting the position of the rotating shaft 1 is attached to the sensor. The gap sensor 4 of the radial magnetic bearing 2 detects the position of the rotating shaft 1 in the radial direction, and the gap sensor 4 of the thrust magnetic bearing 3 detects the position of the rotating shaft 1 in the axial direction. In addition, as this gap sensor, for example, an inductance type displacement sensor that detects displacement by a change in inductance can be used.

Here, as shown in FIG. 1, the magnetic bearings (electromagnets) 2 and 3 and the gap sensor 4 (hereinafter simply referred to as a sensor) are connected to a controller 5. The exciting voltage or exciting current of the electromagnets 2 and 3 is controlled based on. This exciting voltage or exciting current is amplified by the power amplifier 10 and sent to the electromagnet. The controller 5, the power amplifier 10, the magnetic bearings (electromagnets) 2, 3, and the sensor 4 constitute a magnetic bearing control device according to the present invention.

As shown in FIG. 1, the controller 5 includes an A / D converter 6 and a DSP (Digital Signal Process).
or) 7, a D / A converter 8, and a memory 9. DS
P7 controls and processes digital signals.
In cooperation with the programs and data stored in the controller, the arithmetic processing of a signal for controlling the excitation voltage or the excitation current of the electromagnet is performed so as to support the rotating shaft 1 at the target position. The analog displacement signal detected by the gap sensor 4 is input to an A / D converter 6 and converted into a digital signal. Then, the signal digitized by the A / D converter 6 is input to the DSP 7, and arithmetic processing is performed. DS
A digital signal is output from P7 as a result of the arithmetic processing, and this digital signal is converted into an analog signal by the D / A converter 6. The analog signal is supplied to the power amplifier 1
0, the magnetic bearings (electromagnets) 2 and 3 are sent to the coils in the electromagnets as exciting voltages or exciting currents. This exciting voltage or exciting current generates a magnetic attraction between the iron core in the electromagnet and the rotating shaft 1, and controls the position of the rotating shaft 1.

By the way, in order to design and manufacture the above-described magnetic bearing control device, it is necessary to control a rotating shaft,
It is necessary to model (identify) the magnetic bearing system. In the present embodiment, this modeling is performed using a fuzzy neural network. In this fuzzy neural network, an input space is vaguely divided by a fuzzy set, and a multilayer neural network is applied to each of these fuzzy sets. That is, a fuzzy model of the rotating shaft / magnetic bearing system is defined by an if-then type rule, similarly to a generally used fuzzy control rule, and is obtained by a fuzzy neural network having a learning function. Hereinafter, modeling of a rotating shaft / magnetic bearing system using this fuzzy neural network will be described.

First, a fuzzy model is prepared which is defined in the if-then format and describes the dynamic characteristics of the rotating shaft / magnetic bearing system. In the present embodiment, the displacement of the center of the rotating shaft 1 from the reference point is x ₁ , the first-order differential value with respect to the displacement x ₁ of the rotating shaft is x ₂ , the exciting voltage or exciting current of the electromagnet is u,
The displacement of the rotating shaft from the reference point at the center of the rotating shaft 1 after the elapse of the unit time is defined as y. the x _{1, x 2} and antecedent _{variables, x 1,}
Assuming that x ₂ and u are consequent variables, a fuzzy model describing the dynamic characteristics of the rotating shaft / magnetic bearing system is as follows. _{^{L i: if x 1 is A}} 1 i, x 2 is A 2 i then y i
^{_{= A 1 i x 1 + a}} 2 i x 2 + a 3 i u L i denotes the i-th fuzzy rule, during this fuzzy rule _L, A 1, _{A 2} is a fuzzy set (antecedent parameter), _{a 1} , A ₂ , a ₃ indicate consequent parameters.

Here, in the case of the radial magnetic bearing 2, the "reference point" indicates the point O shown in FIG. 2 (a longitudinal sectional view including the radial magnetic bearing 2 in FIG. 1). That is, it is the intersection of the line connecting the pair of bearings 2a and 2b facing in the horizontal direction and the line connecting the pair of bearings 2c and 2d facing in the vertical direction. It is ideal for control that the center of the rotating shaft 1 coincides with the reference point O, and the controller 5 controls the rotating shaft 1 of the magnetic bearings 2 and 3.
The exciting voltage or exciting current of the electromagnet is controlled so that the center coincides with the reference point O. The “reference point” in the case of the thrust magnetic bearing 3 is a middle point of a line connecting the pair of axially opposed thrust magnetic bearings 3 shown in FIG. The term “displacement” means a displacement of the center of the rotation axis 1 from the reference point O, unless otherwise indicated.

Here, if the antecedent parameter and the consequent parameter in the fuzzy rule L can be determined, the modeling of the rotating shaft / magnetic bearing system can be realized. As described above, the optimal solutions of various parameters of the fuzzy rule L are obtained by a fuzzy neural network having a learning function. This enables more accurate modeling.

FIG. 3 is a configuration diagram of a fuzzy neural network formed based on the fuzzy rule L. The configuration of the fuzzy neural network is
Depending on the number of antecedent parameters in the fuzzy rule L and the configuration of the consequent part, a configuration different from the configuration shown in FIG. 3 can be adopted. In this fuzzy neural network, input / output data obtained from experiments, that is, displacement of the rotating shaft at a specific point in time, first-order differential value with respect to the displacement of the rotating shaft, exciting voltage or exciting current of the electromagnet,
Learning is performed based on the displacement of the rotating shaft after the elapse of the unit time.

Firstly, the fuzzy neural network of FIG. 3, the displacement x ₁ of the rotation shaft at a particular point in time obtained in the _experiment, first-order differential value x ₂ with respect to the displacement x ₁ of the rotation _axis, the excitation voltage or the excitation current of the electromagnet Enter u. Based on these input values, a calculation is performed in consideration of a predetermined coupling load w _i (i = 1, 2,..., 11), and as a calculation result, a displacement y of the rotating shaft after a unit time has elapsed is estimated (calculated). ) Is done.

Next, the difference between the estimated displacement y of the rotating shaft after the lapse of the unit time and the actual displacement of the rotating shaft after the lapse of the unit time obtained by the experiment is determined, and this difference is minimized. Is the connection weight w _i (i =
Modify 1,2, ..., 11) as appropriate.

As described above, the input of the data obtained by the experiment, the comparison between the calculated value and the actual value, and the connection weight w _i (i = 1, 2,
, 11) is repeated a predetermined number of times, it is possible to obtain an optimum coupling load w _i (i = 1, 2,..., 11). When the optimum coupling load w _i (i = 1, 2,..., 11) is obtained, the respective parameters of the fuzzy rule L are determined based on the calculated values, and the fuzzy rule L is used for the rotation shaft / magnetic bearing system. Dynamic characteristics can be described. This allows accurate modeling.

When the control object is modeled, a control device for operating the modeled control object (rotary shaft / magnetic bearing system), that is, a magnetic bearing control device can be designed. In the present embodiment, the exciting voltage or the exciting current of the electromagnet is controlled by the sliding mode control, and a control system with high robustness is realized.

[0025] In designing the sliding mode control, first, as a state variable X, to define the deviation [delta] ₁ and the differential value [delta] ₂ of the deviation [delta] ₁ and the target value of the displacement of the rotary shaft and the displacement of the rotating shaft. That is, the state variable X is defined as follows. X = [δ ₁ δ ₂ ] ^T

The switching function σ (X) = 0 used in the sliding mode control is defined as follows. σ (X) = SX S is a matrix matrix, which can be determined by a general sliding mode control switching setting method.

Further, the exciting voltage or exciting current u of the electromagnet, which is the output from the controller 5, is defined as state feedback as follows. u = KX where K is a state feedback gain vector.

Here, the state feedback gain vector K is determined so as to satisfy the following expression which is a condition for existence of the sliding mode.

(Equation 1) Thus, the state variable X is constrained by the switching plane, and a stable control system with high robustness, that is, a sliding mode control system can be designed.

The sliding mode control system designed as described above is realized by the DSP 7 of the controller 5. Hereinafter, the sliding mode control by the DSP 7 will be described in detail. FIG. 4 is a block diagram illustrating a configuration inside the DSP 7 of the controller 5 according to the present embodiment. As shown in the figure, a digital low-pass filter (LPF) 71 and a sliding mode control unit 72 are provided in the DSP 7. In this figure, x is the displacement of the rotating shaft 1 measured by the sensor 4 of the magnetic bearing, r is the target value of the displacement of the rotating shaft 1, δ ₁ is the deviation between r and x, and δ ₂ is the first order of δ ₁ Differential value.

When the displacement x of the rotary shaft 1 is input from the sensors 4 of the magnetic bearings 2 and 3 to the DSP 7 via the A / D converter 6, the difference value 73 sets the target value r of the displacement of the rotary shaft 1 and the displacement. deviation [delta] ₁ and x is determined. The above-mentioned δ ₁ is input to the low-pass filter 71, and by passing through the low-pass filter 71, high-frequency components that can be ignored for control are cut. In this way, higher-performance control is realized.

The signal (deviation δ ₁ ) that has passed through the low-pass filter 71 is input to a sliding mode control section 72. The deviation δ ₁ input to the sliding mode control unit 72 is
The signals are input to the first gain means 74 and the differentiating means 75, respectively. Deviation δ input to differentiating means 75
₁ is differentiated, the differential value [delta] ₂ is determined. Then, the differential value δ ₂ is input to the second gain means 76.

Here, the input to the first gain means 74 is
Deviation δ₁And the differential value δ₂Suitable gain depending on
K_1aOr K_1bSwitch SW1 for selecting
ing. This selected gain K_1aOr K_1bIs on
This is the state feedback gain described above. First gainer
Stage 74 has an appropriate gain K by switch SW1._{1 a}
Or K_1bAnd the above deviation δ₁Multiply this gain by
Match. Also, the second gain means 76 has the deviation δ
₁And the differential value δ of the input deviation ₂Suitable gay depending on
K_2aOr K_2bSwitch SW2 for selecting
The selected gain is also the same as the state
Feedback gain. The second gain means 76 includes a switch.
Appropriate gain K by switch SW2_2aOr K_2bChoose
And the differential value δ of the above deviation₂Multiplied by this gain
You.

The sliding mode control unit 72
The deviation δ ₁ multiplied by an appropriate gain by the first gain unit 74 and the deviation δ ₂ multiplied by an appropriate gain by the second gain unit
Is output to the D / A conversion unit 8 of the controller 5 as a calculation result.

Thus, in the state space, the switching plane σ
By providing (X) and switching an appropriate gain according to the deviation δ ₁ and the differential value δ _{2 of the} deviation, the state can be constrained to this switching plane σ (X), and the above-described sliding mode control is realized. Is done. Thereby, highly robust control can be performed.

Thus, the deviation δ ₁ and the differential value δ of the deviation
_The above-described sliding mode control is realized by selecting appropriate gains according to ₂ and multiplying and adding them. Thereby, highly robust control can be performed.

Here, the results of a floating / rotation experiment of the rotating shaft 1 performed by installing the above-described magnetic bearing control device on a vertically placed ET450 type pump are shown in FIGS. In addition,
The results shown in FIGS. 5 to 9 relate to a pair of magnetic bearings facing in the horizontal direction.

FIG. 5A is a graph showing the relationship between the elapsed time t (ms) and the excitation voltage u (V) of the electromagnet, and FIG.
Is a graph showing the relationship between the elapsed time t (ms) and the deviation δ ₁ (μm). As shown in this figure, it can be seen that the rotating shaft 1 is quickly (within 100 ms) from the touch-down state (t = 0) and is in a stable floating state without overshoot.

FIG. 6 is a graph showing a control result when a disturbance is applied in the plus (+) direction in the stable levitation state of FIG. 5, and FIG. 7 is a minus (-) in the stable levitation state of FIG. 9 is a graph showing a control result when a disturbance is applied in a direction. (A) of these figures shows the elapsed time t (ms)
And the excitation voltage u (V) of the electromagnet, and (b) shows the relationship between the elapsed time t (ms) and the deviation δ ₁ (μm). From these figures, it can be seen that the control is performed with good convergence even when a disturbance is applied in any of the plus (+) and minus (-) directions.

FIG. 8 shows the rotational speed (mi) when the rotational speed is increased from the stable floating state shown in FIG. 5 to the rated rotational motion.
3 is a graph three-dimensionally showing a relationship among n ⁻¹ ), vibration frequency (Hz), and amplitude (μm). On the other hand, FIG. 9 shows when the rotation speed is reduced from the rated rotation motion to the stable floating state,
It is the graph which showed the rotation speed (min- ¹ ), the vibration frequency (Hz), and the relationship of the amplitude (micrometer) three-dimensionally. In these figures, the X axis is the vibration frequency, the Y axis is the rotational speed,
The Z axis indicates amplitude. From these figures, it can be seen that there is only a component synchronized with the rotation and the vibration such as the asynchronous vibration component is considerably small in both the case of the speed increase and the case of the deceleration.

[0040]

As described above, according to the present invention, since the rotating shaft / magnetic bearing system can be controlled by the sliding mode control, it is possible to perform control with high robustness. In addition, when the control system of the rotating shaft / magnetic bearing is modified / adjusted, there is no need to change the hardware, and the control system of the rotating shaft / magnetic bearing can be easily modified / adjusted.

Furthermore, since an accurate model of a rotating shaft / magnetic bearing system can be obtained by a fuzzy neural network, control based on an accurate model of a rotating shaft / magnetic bearing system can be performed, and input / output data obtained by experiments can be obtained. , And at the same time, it is possible to control the rotating shaft / magnetic bearing system with high robustness.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of a magnetic bearing according to an embodiment of the present invention.

FIG. 2 is a longitudinal sectional view including the radial magnetic bearing of FIG. 1;

FIG. 3 is a configuration diagram of a fuzzy neural network configured based on fuzzy rules in one embodiment of the present invention.

FIG. 4 illustrates a controller D according to an embodiment of the present invention.
FIG. 3 is a block diagram illustrating a configuration in an SP.

5A is a graph showing a relationship between an elapsed time t (ms) and an excitation voltage u (V) of an electromagnet, and FIG.
6 is a graph showing a relationship between (ms) and a deviation δ ₁ (μm).

6A is a graph showing the relationship between the elapsed time t (ms) and the excitation voltage u (V) of the electromagnet when disturbance is applied in the plus (+) direction in the stable levitation state of FIG. 5, and FIG. )
Is a graph showing the relationship between the elapsed time t (ms) and the deviation δ ₁ (μm) at that time.

7 is a graph showing the relationship between the elapsed time t (ms) and the excitation voltage u (V) of the electromagnet when disturbance is applied in the minus (-) direction in the stable levitation state in FIG. 5, and FIG. 5 is a graph showing a relationship between elapsed time t (ms) and deviation δ ₁ (μm).

8 is a three-dimensional view showing the relationship among the rotation speed (min ⁻¹ ), vibration frequency (Hz), and amplitude (μm) when the rotation speed is increased from the stable levitation state in FIG. 5 to the rated rotation motion. It is a graph.

FIG. 9 shows a three-dimensional relationship between the rotation speed (min ⁻¹ ), the vibration frequency (Hz), and the amplitude (μm) when the rotation speed is reduced from the rated rotation motion in FIG. 5 to the stable floating state. It is a graph.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 Rotary shaft 2 Radial magnetic bearing 3 Thrust magnetic bearing 4 Sensor 5 Controller 7 DSP 72 Sliding mode control part 73 Difference means 74 First gain means 75 Differentiation means 76 Second gain means 77 Addition means K _1a , K _1b , K _2a , K _2b gain SW1, SW2 switch

Claims (2)

[Claims] A reference point is set in a magnetic bearing that supports a rotating shaft by a magnetic force of an electromagnet, and a sensor for detecting displacement of the rotating shaft from the reference point is provided. A magnetic bearing control device that controls a current or a voltage supplied to the electromagnet based on the displacement, and a difference unit that calculates a deviation between the displacement of the rotating shaft detected by the sensor and a target value of the displacement of the rotating shaft. A differential means for obtaining a differential value of the deviation obtained by the differential means; and a gain determined based on a fuzzy model of a predetermined rotating shaft / magnetic bearing system, the deviation obtained by the differential means and First gain means for selecting according to the differential value of the deviation obtained by the differentiating means and multiplying by the deviation; and a fuzzy module for a predetermined rotating shaft / magnetic bearing system. A second gain means for selecting a gain determined based on the differential value in accordance with the difference obtained by the difference means and the differential value of the difference obtained by the differentiating means, and multiplying the gain by the differential value of the difference. And an adding unit that outputs a value obtained by adding the value calculated by the first gain unit and the value calculated by the second gain unit as a current or a voltage to be supplied to the electromagnet. Magnetic bearing control device. 2. A fuzzy neural system for a fuzzy model of the rotating shaft / magnetic bearing system, which is described including a displacement of the rotating shaft, a differential value of the displacement of the rotating shaft, and a current or voltage supplied to the electromagnet. By forming a network, and inputting the displacement of the rotating shaft, the differential value of the displacement of the rotating shaft, and the current or voltage supplied to the electromagnet to the fuzzy neural network, the displacement of the rotating shaft after the elapse of a unit time is obtained. Guess, find the difference between the estimated displacement of the rotation axis after the elapse of the unit time and the actual displacement of the rotation axis after the elapse of the unit time, and repeatedly train the fuzzy neural network to minimize the difference. Based on the learned result, an accurate fuzzy model of the rotating shaft / magnetic bearing system is derived, and the obtained fuzzy model of the rotating shaft / magnetic bearing system is The magnetic bearing control device according to claim 1, wherein the gain is determined.