JP6269080B2 - Magnetic bearing device and vacuum pump - Google Patents

Description

  The present invention relates to a magnetic bearing device and a vacuum pump.

  In a magnetic bearing mounting apparatus such as a turbo molecular pump, an actuator is used based on a deviation amount between the rotor's flying displacement and the target position in order to keep the rotor, ie, the rotor, floating at a predetermined target position without contacting the stator. The magnetic attraction force of the electromagnet, that is, the excitation current is feedback controlled in real time.

  In the detection of rotor levitation displacement, the type that is detected by a dedicated sensor has been the mainstream. In recent years, in order to reduce the size, reduce the price, and improve reliability, a sensorless type that omits a dedicated sensor and combines an electromagnet that generates levitation control force with a conventional actuator function as well as a sensing function, that is, Self-sensing type is being put into practical use.

  In either type, the inductance method is employed as a sensing function. In the sensorless type, an inductance method is a change in inductance caused by a change in a floating gap, that is, a gap between an electromagnetic coil and a shaft, by applying a high frequency carrier wave, that is, a sensor carrier, to an electromagnetic coil (hereinafter also simply referred to as a coil). The sensor carrier is amplitude-modulated and demodulated to obtain a floating gap (displacement) signal. In the demodulation processing, a direct method has been proposed in which a modulated wave signal is synchronously sampled from an AD converter by applying a digital technique, and smoothing processing that causes a delay is unnecessary (for example, Patent Document 1). .

  In general, in a turbo molecular pump, there are five control axes, ie, an upper X axis, an upper Y axis, a lower X axis, a lower Y axis, and a thrust Z axis. An M-side electromagnet is provided. The currents flowing in the coils of the respective electromagnets can be divided into components by function, and the DC force used for biasing for balancing the gravity force acting on the rotator unit R, improving the linearity of the levitation force, and displacement sensing. It is composed of a bias current ib, a control current ic for floating the rotating body to a predetermined position, and a current is of a sensor carrier component for position detection.

  It is known that two errors occur in obtaining a displacement signal from the current component flowing in the above-described electromagnet coil. One is an error (hereinafter referred to as a first error) in which the control current component ic that remains additively in the modulated wave signal before the demodulation process is mixed as a displacement signal during the demodulation process. The other is an error of the displacement signal (hereinafter referred to as a second error) caused by the non-linearity of the magnetic characteristics of the electromagnet core, that is, the electromagnet core. The second error becomes prominent when the current flowing through the electromagnetic coil increases. In the following, the problem of removing the first error is called a first problem, and the problem of removing the second error is called a second problem. A signal obtained by removing the first error and the second error from the displacement signal having an error is particularly referred to as a true estimated displacement signal.

  In Patent Document 1, the first problem is solved by making the magnitudes of the control currents ic of the two opposing electromagnets facing each other in the magnetic bearing device equal and opposite in sign. However, in this case, it is necessary to provide a narrow limit of about | ic | <ib in the change range of the control current ic. Due to this limitation, the total current flowing through the coil of the electromagnet is set small. That is, since Patent Document 1 does not assume that a large current flows through the coil of the electromagnet, the second error does not become significant and the second problem does not become a problem.

  The turbo molecular pump attached to the vacuum chamber is mounted not only in the vertical direction, that is, the axis of the rotating shaft of the rotating body extends in the vertical direction but also in the horizontal direction, that is, the axis of the rotating shaft of the rotating body. May be attached extending in a direction perpendicular to the vertical direction. In the case of the horizontal orientation, specifically, it is necessary to flow a large control current ic only to one of the opposing electromagnet coils positioned above the rotating body. Thus, the magnetic bearing device of the turbo molecular pump must be controlled according to the direction in which the turbo molecular pump is attached.

  In addition, when an external vibration such as an earthquake is applied, the rotating body is transiently displaced from the predetermined floating position, but the rotating body of the opposing electromagnet coils moves away to return to the predetermined floating position accordingly. On the other hand, a control signal ic having a large amplitude needs to be supplied to only one of them.

  As described above, in the case where a large current needs to flow only in one of the facing electromagnet coils, the invention described in Patent Document 1 cannot solve the first problem and the second problem.

JP 2001-177919 A

  When the asymmetric current described above is applied, the first problem can be solved by the invention described in Japanese Patent Application No. 2013-021681. However, the second problem is not solved.

  That is, even when a large current is passed through only one of the opposing electromagnet coils, it is necessary to remove an error in the displacement signal resulting from the non-linearity of the magnetic characteristics of the electromagnet core.

(1) A magnetic bearing device according to a preferred embodiment of the present invention is provided on each of a rotating body and a plurality of control shafts for controlling the floating of the rotating body, and is disposed to face each of the control shafts via the rotating body. The voltage applied to each of the first electromagnet and the second electromagnet and the first electromagnet and the second electromagnet is PWM-controlled, and an electromagnet current superimposed with a sensor carrier signal for detecting the floating position of the rotating body is supplied. An excitation amplifier provided in each of the first electromagnet and the second electromagnet, a current sensor provided in each of the excitation amplifiers to detect an electromagnet current, an estimated displacement signal based on the electromagnet current, and an estimated displacement signal and a control unit for PWM controlling the excitation amplifier based on the control unit, in obtaining an estimated displacement signals in each control axes of the control shaft, the flow to the first electromagnet That the detection signal of the electromagnet current (hereinafter, the first current signal), the detection signal of the electromagnet current flowing through the second electromagnet (hereinafter, second current signal), and, combined by adding the first current signal and a second current signal the resulting sum signal performs AD sampling in synchronization with the sensor carrier signals, respectively, to generate a demodulated displacement signals by AD sampled sum signal in synchronization with the sensor carrier signals, due to non-linearity of the magnetic characteristics of the first electromagnet The first correction signal for reducing the influence on the demodulated displacement signal is generated by arithmetic processing based on the non-linearity of the first current signal after AD sampling and the magnetic characteristics of the first electromagnet, and the magnetism of the second electromagnet The second correction signal for reducing the influence on the demodulated displacement signal due to the non-linearity of the characteristics is the second current signal after AD sampling and the magnetic characteristics of the second electromagnet. Generated by the arithmetic processing based on linearity, by a third correction signal as a difference between the first correction signal and the second correction signal generated by the arithmetic processing, subtracting the third correction signal from the demodulated displacement signal, estimate A displacement signal is obtained.
(2) In a further preferred embodiment, the sum signal satisfies fc = (n + 1/2) · fs (where n is an integer of 0 or more), where fc is the frequency of the sensor carrier signal. AD sampling is performed at a sampling frequency fs, a timing near the maximum peak position of the sensor carrier signal, and a timing near the minimum peak position, and the data value of the sum signal sampled near the maximum peak position is d1, When the data value of the sum signal sampled in the vicinity of the minimum peak position is d2, the value d3 calculated by d3 = (d1−d2) / 2 is set as the demodulated displacement signal .
(3) In a further preferred embodiment, the rotating body has a movable boundary, the first correction signal is composed of a function (hereinafter referred to as a first function) made up of the first current signal, and the second correction signal is made up of the second correction signal. When an adjustment (hereinafter referred to as a first coefficient adjustment) is made up of a function consisting of a current signal (hereinafter referred to as a second function) and a coefficient used for the first function is optimized (hereinafter referred to as a first coefficient adjustment), the rotating body is placed on the first electromagnet side. The first coefficient adjustment is performed after the rotating body is moved to the movable boundary of the rotating body and the rotating body is magnetically attracted and fixed at a predetermined bias current value, and the coefficient used for the second function is optimized. When the second coefficient adjustment is performed, the rotating body is moved to the movable boundary of the rotating body on the second electromagnet side, and the rotating body is magnetically attracted and fixed at a predetermined bias current value. The second coefficient adjustment is performed.
(4) In a further preferred embodiment, the first coefficient adjustment is performed by superimposing an applied signal having a predetermined frequency on at least the electromagnet current on the first electromagnet side, and then the amplitude of the estimated displacement signal having a predetermined frequency. The second coefficient adjustment is performed by superimposing an applied signal having a predetermined frequency on at least the electromagnet current on the second electromagnet side, and then estimating a displacement having a predetermined frequency. The present invention is characterized in that the signal amplitude is less than a predetermined level.
(5) In a more preferred embodiment, a plurality of predetermined bias current values are provided, and the first coefficient adjustment and the second coefficient adjustment are performed at each of the plurality of predetermined bias current values. To do.
(6) In a more preferred embodiment, the first coefficient adjustment and the second coefficient adjustment are performed a plurality of times at a predetermined bias current value.
(7) In a further preferred embodiment, the third correction signal is generated by calculating a difference between the first correction signal and the second correction signal at a predetermined ratio, and the predetermined ratio is determined by the time when the rotating body is time-dependent. The magnetic attraction force when the applied signal having the predetermined frequency is superimposed on the electromagnet current flowing in the first electromagnet and the electromagnet current flowing in the second electromagnet in the state where the magnetic levitation control is performed so as not to change the position. wherein Ru is set so that the magnetic attraction force in the case of superimposing the applied signal having a predetermined frequency is canceled out.
The vacuum pump according to a preferred embodiment of (8) The present invention, a pump rotor exhaust function portion is formed, a motor for rotating the pump rotor, magnetically levitated rotating member of the pump rotor according to claim 1 to 7 And a magnetic bearing device according to any one of the above.

  According to the present invention, it is possible to maintain a good S / N ratio with respect to the displacement signal of the rotating body even when a large current is passed through only one of the opposing electromagnet coils. Furthermore, the vibration of the pump casing can be reduced, and the floating control performance of the rotating body can be improved.

The figure which shows schematic structure of the pump unit 1 of a vacuum pump. The block diagram which shows schematic structure of a control unit. The schematic diagram which shows the magnetic bearing electromagnet 45 for 1 axis | shaft of the control shaft with which the magnetic bearing was equipped. The figure which shows the structure of the excitation amplifier 43 provided in each magnetic bearing electromagnet 45. FIG. The figure which shows the applied voltage to the electromagnet coil by the excitation amplifier 43, and the electric current which flows into an electromagnet coil. The figure which showed the shaft 5 and the electromagnet 45 when the pump unit 1 is arrange | positioned sideways. The figure which shows the control signal dependence of the electric current which flows into the electromagnet coil 455 in the current limit circuits 408p and 408m. The functional block diagram for one axis of magnetic bearing control in the control part 44. FIG. The figure which showed the reverse quantity of (a) BH curve of the laminated silicon steel plate of an electromagnet core, and (b) relative magnetic permeability. The figure which showed the position of the shaft 5 of the rotary body unit R at the time of adjustment. The figure which showed operation | movement of the correction process calculating part 415 at the time of adjustment. The figure which showed operation | movement of the correction process calculating part 415 at the time of normal control.

  Hereinafter, embodiments for carrying out the present invention will be described with reference to the drawings. FIG. 1 is a diagram for explaining a schematic configuration of a vacuum pump according to the present embodiment, and shows a cross-sectional configuration of a pump unit 1 of a magnetic levitation turbomolecular pump. The turbo molecular pump includes a pump unit 1 shown in FIG. 1 and a control unit (not shown) that drives the pump unit 1.

  The pump unit 1 has a turbo pump part composed of the rotary blades 4a and the fixed blades 62, and a drag pump part composed of the cylindrical part 4b and the screw stator 64, that is, a thread groove pump. Here, a screw groove is formed on the screw stator 64 side, but a screw groove may be formed on the cylindrical portion 4b side. The rotary blade 4a and the cylindrical part 4b, which are the rotation side exhaust function part, are formed in the pump rotor 4. The pump rotor 4, the shaft 5, and the rotor disk 55 are fastened to each other to constitute a rotating body unit R.

  The plurality of stages of fixed blades 62 are alternately arranged with the rotary blades 4a in the axial direction. Each fixed wing 62 is placed on the base 60 via the spacer ring 63. When the fixing flange 61c of the pump casing 61 is fixed to the base 60 with a bolt, the stacked spacer ring 63 is sandwiched between the base 60 and the locking portion 61b of the pump casing 61, and the fixed blade 62 is positioned.

  The shaft 5 is supported in a non-contact manner by magnetic bearings 67, 68 and 69 provided on the base 60. As will be described later, the magnetic bearings 67, 68, and 69 are self-sensing magnetic bearings that estimate a change in the flying position based on an electromagnet current on which a sensor carrier component is superimposed. Note that the electromagnets constituting the axial magnetic bearing 69 are arranged so as to sandwich the rotor disk 55 provided at the lower end of the shaft 5 in the axial direction. The shaft 5 is rotationally driven by a motor 42.

  The motor 42 is a synchronous motor, and a DC brushless motor is used in the present embodiment. The motor 42 includes a motor stator 42 a disposed on the base 60 and a motor rotor 42 b provided on the shaft 5. The motor rotor 42b is provided with a permanent magnet. When the magnetic bearing is not operating, the shaft 5 is supported by emergency touchdown bearings 66a and 66b.

  An exhaust port 65 is provided at the exhaust port 60 a of the base 60, and a back pump is connected to the exhaust port 65. By rotating the rotating unit R at high speed by the motor 42 while magnetically levitating, the gas molecules on the intake port 61a side are exhausted to the exhaust port 65 side.

  FIG. 2 is a block diagram showing a schematic configuration of the control unit. The AC input from the outside is converted from alternating current to direct current by a DC power supply 40 provided in the control unit. The DC power source 40 generates a power source for the inverter 41, a power source for the excitation amplifier 43, and a power source for the control unit 44, respectively.

The inverter 41 that supplies current to the motor 42 is provided with a plurality of switching elements. The motor 42 is driven by controlling the on / off of these switching elements by the control unit 44.

  The ten magnetic bearing electromagnets 45 (hereinafter also referred to as electromagnets 45) shown in FIG. 2 are magnetic bearing electromagnets provided in the magnetic bearings 67, 68, and 69, respectively. The magnetic bearing used in the turbo molecular pump shown in FIG. 1 is a five-axis control type magnetic bearing, and the radial magnetic bearings 67 and 68 are two-axis magnetic bearings, each having two pairs, that is, Four magnetic bearing electromagnets 45 are provided. The axial magnetic bearing 69 is a uniaxial magnetic bearing and includes a pair, that is, two magnetic bearing electromagnets 45. Excitation amplifiers 43 for supplying current to the magnetic bearing electromagnets 45 are provided in each of the ten magnetic bearing electromagnets 45, and the control unit includes a total of ten excitation amplifiers 43.

  The control unit 44 that controls the driving of the motor 42 and the driving of the magnetic bearing is configured by a digital arithmetic unit such as an FPGA (Field Programmable Gate Array) and its peripheral circuits, for example. The control unit 44 outputs a PWM control signal 441 for on / off control of a plurality of switching elements provided in the inverter 41 to the inverter 41, and each excitation amplifier 43 to each excitation amplifier 43. PWM gate drive signals 443 for ON / OFF control of the switching elements included in the output are respectively output. Further, as will be described later, a signal 442 related to a phase voltage and a phase current related to the motor 42 and an electromagnet current signal 444 related to a magnetic bearing are input to the control unit 44.

  FIG. 3 is a schematic diagram showing the magnetic bearing electromagnet 45 for one control shaft provided in the magnetic bearings 67 and 68. The magnetic bearing electromagnet 45 is configured by winding an electromagnet coil 455 around an electromagnet core 450. Two magnetic bearing electromagnets 45 are opposed to each other so as to sandwich the floating central axis, that is, the floating target position J. As described above, the excitation amplifier 43 is provided for each magnetic bearing electromagnet 45. In FIG. 3, the displacement d that approaches the P-side magnetic bearing electromagnet 45 on the right side of the drawing is positive. The magnetic bearing electromagnet 45 whose displacement is negative will be referred to as the M-side magnetic bearing electromagnet 45.

(Description of electromagnet currents Ip, Im and sum signal Ip + Im)
In the 5-axis control type magnetic bearing in the present embodiment, the electromagnet current of each magnetic bearing electromagnet 45 is divided into components by function, and bias current ib, levitation control current ic, and position detection sensor carrier component current. is is included. When the current flowing through the P-side magnetic bearing electromagnet 45 is Ip and the current flowing through the M-side magnetic bearing electromagnet 45 is Im, the following equation (1) is obtained. isp is a sensor carrier component on the P side, and ism is a sensor carrier component on the M side. However, the amplitudes of isp and ism have opposite signs.
Ip = ib + ic + isp
Im = ib-ic + ism (1)

  The bias current ib is a direct current or an extremely low frequency band, and is used as a bias for a balance force with gravity acting on the rotator unit R, a linearity improvement of the levitation force, and a displacement sensing.

  The levitation control current ic is a control current that levitates the shaft 5, that is, the rotating body unit R to a predetermined position. Since the levitation control current ic changes according to the fluctuation of the levitation position, the frequency band is from DC to about 1 kHz.

  The sensor carrier component is is a current component used for detecting the floating position displacement of the shaft 5, that is, the floating position displacement of the rotating body unit R. For the sensor carrier component is, a frequency in a frequency band of several kHz to several tens of kHz (1 kHz << fc << 100 kHz) is usually used in order to suppress the influence on the levitation control current as much as possible.

  Generally, in a magnetic bearing for industrial use, a voltage control type PWM amplifier is used as the excitation amplifier 43. That is, the electromagnet current is controlled by controlling the voltage applied to the electromagnet coil of the magnetic bearing electromagnet 45.

Of the voltages Vp and Vm applied to the electromagnet coil, the sensor carrier components vsp and vsm are applied in opposite phases, and therefore are expressed by the following equation (2). However, ωc = 2πfc, and fc is the sensor carrier frequency. T is time, and v is a constant amplitude value.
vsp = −v × sin (ωc × t)
vsm = v × sin (ωc × t) (2)

By the way, since the gap between the magnetic bearing electromagnet 45 and the shaft 5 (see FIG. 3) and the inductance of the electromagnet coil are inversely proportional, the inductances Lp and Lm of the P-side electromagnet coil and the M-side electromagnet coil are expressed by the following formula (3 ) Holds. Here, D is a gap when the shaft 5 is at the levitation center axis, that is, at the levitation target position, and d is a displacement from the levitation target position. A is a constant.
1 / Lp = A × (D−d)
1 / Lm = A * (D + d) (3)
In the present invention, in order to solve the second problem, equation (3) is corrected as equation (3a) described later.

Regarding the sensor carrier component, there is a relationship as shown in the following equation (4) between the voltage applied to the electromagnetic coil and the current flowing through the electromagnetic coil. However, coil resistance was ignored.
vsp = Lp × d (isp) / dt
vsm = Lm × d (ism) / dt (4)

From the expressions (2), (3), and (4) described above, the sensor carrier components isp and ism of the current flowing through the electromagnetic coil are expressed as the following expression (5). Note that B = v × A / ωc. Thus, the sensor carrier components isp and ism are amplitude-modulated by the time change of the displacement d. On the other hand, since the bias current ib and the levitation control current ic have low frequencies, the influence of the displacement fluctuation can be ignored.
isp = −v × sin (ωc × t−π / 2) / (ωc × Lp)
= −B (D−d) × sin (ωc × t−π / 2)
ism = v × sin (ωc × t−π / 2) / (ωc × Lm)
= B (D + d) × sin (ωc × t−π / 2) (5)
In the present invention, in order to solve the second problem, equation (5) is corrected as equation (5a) described later.

To summarize the above results, information on the displacement d can be obtained by detecting the sensor carrier components isp and ism. The total currents Ip and Im flowing through the P-side and M-side magnetic bearing electromagnets 45 are expressed by the following equation (6).
Ip = ib + ic−B (D−d) × sin (ωc × t−π / 2)
Im = ib−ic + B (D + d) × sin (ωc × t−π / 2) (6)
In the present invention, in order to solve the second problem, equation (6) is corrected as equation (6b) described later.

(Description of two-quadrant excitation amplifier)
FIG. 4 is a diagram showing a configuration of the excitation amplifier 43 provided corresponding to each magnetic bearing electromagnet 45. The excitation amplifier 43 is obtained by further connecting two series-connected switching elements and diodes in parallel. The magnetic bearing electromagnet 45 is connected between the middle of the switching element SW10 and the diode D10 and the middle of the switching element SW11 and the diode D11.

  The switching elements SW10 and SW11 receive a PWM control signal for controlling the bias current ib, the levitation control current ic, and the sensor carrier component is (a PWM gate drive signal in FIG. 2) as a gate signal (gate drive voltage) from the control unit 44. 443) is input. The switching elements SW10 and SW11 are simultaneously turned on and off. When both are on, currents (currents Ip and Im described above) flow as indicated by solid arrows, and when both are off, currents (as indicated by dashed arrows) The above-described currents Ip and Im) flow. The on-state current value is measured by the current sensor 101A, and the off-state current value is measured by the current sensor 101B. For example, shunt resistors are used for the current sensors 101A and 101B, and the voltage of the shunt resistor is used as a current detection signal. The current detection signal is input to the control unit 44.

  FIG. 5 is a diagram illustrating an example of an applied voltage (line L1) to the electromagnet coil by the excitation amplifier 43 and a current (line L2) flowing through the electromagnet coil. When the two switching elements SW10 and SW11 are turned on, a voltage is applied to the electromagnet coil and the current increases. Further, when the switching elements SW10 and SW11 are turned off, the reverse voltage is applied to the electromagnetic coil due to the conduction of the diodes D10 and D11, and the current is reduced. Therefore, the current line L2 shows both an increase and a decrease in current in one PWM carrier cycle and a longer sinusoidal change. This sinusoidal change corresponds to a change in the sensor carrier component.

  Here, for the following reasons, when considering the first problem and the second problem, Expression (6) described above is transformed into Expression (6a) described later. As will be described later, Equation (6) is also transformed into Equation (6b) in order to derive a correction signal for removing the second error. FIG. 6 shows, as an example, the shaft 5 of the rotating body unit R and the electromagnet 45 of the magnetic bearing 67 when the pump unit 1 is in the horizontal direction, that is, when the rotating shaft of the rotating body unit R is disposed at right angles to the direction of gravity. Yes. In FIG. 6, the electromagnet 45 of the magnetic bearing 67 has been described, but the same applies to the electromagnet 45 of the magnetic bearing 68. Here, the reference numerals of the electromagnets 45 of the magnetic bearing 67 are reassigned. The M-side electromagnet 45 at the upper right in the drawing is an electromagnet 451m, and the P-side electromagnet 45 at the lower left in the drawing facing the electromagnet 451m is an electromagnet 451p. Similarly, the M-side electromagnet 45 at the upper left in the drawing is an electromagnet 452m, and the P-side electromagnet 45 at the lower right in the drawing facing the electromagnet 452m is an electromagnet 452p.

  The electromagnet 451m and the electromagnet 452m are located above the shaft 5 of the rotating body unit R. Therefore, a large current flows through the electromagnet coils 455 of the electromagnets 451m and 452m in order to counter the gravity applied to the shaft 5, that is, to support the shaft 5 by magnetic attraction. On the other hand, the electromagnet 451 p and the electromagnet 452 p are located below the shaft 5. Therefore, although it is not necessary to support the shaft 5, as will be described later, a small current called a limit direct current i_limit flows through the electromagnet coils 455 of the electromagnet 451p and the electromagnet 452p.

FIG. 7 shows the control signal dependence of the current flowing through the electromagnetic coil 455 in the current limit circuits 408p and 408m shown in FIG. FIG. 7A is a diagram regarding the current flowing through the P-side electromagnet coil 455, and FIG. 7B is a diagram regarding the current flowing through the M-side electromagnet coil 455. As shown in FIG. 6 described above, these currents greatly deviate to the right or left of FIG. 7 when the pump unit 1 is disposed sideways. FIG. 7 is a diagram adapted to the case shown in FIG. As shown in FIG. 7, a large current including a large DC control current ic, which is several times the bias current ib, flows through the M-side electromagnet coil 455 through which a large current flows. On the other hand, a direct current smaller than the bias current ib flows through the opposing P-side electromagnet coil 455, and reaches a limit direct current i_limit necessary for superimposing the sensor carrier.

  Although the case where the pump unit 1 is disposed sideways has been described so far, there is a case where a large current is supplied only to one of the opposing electromagnetic coils 455. For example, it is a case where vibration is applied from the outside such as an earthquake. In this case, the rotator is transiently displaced greatly from the predetermined levitation position, but in order to return to the predetermined levitation position, a large amplitude control current is applied only to one of the opposing electromagnet coils 455 where the rotator has moved away. I need to stream ic. Since external vibrations such as earthquakes can occur in various directions, in this case, in any of the magnetic bearings 67, 68, and 69, a large current can flow only to one of the opposing electromagnet coils 455. There is sex.

From the above, the present invention also assumes that a large current flows only through one of the opposing electromagnet coils 455 in the magnetic bearings of each axis.
As a result, the following two things must be considered. First, the magnitude of the control current ic differs between the P-side electromagnet coil 455 and the M-side electromagnet coil 455. Therefore, the control currents flowing through the P-side electromagnet coil 455 and the M-side electromagnet coil 455 need to be expressed as icp and icm, respectively. As will be described later, the cause of the first error is that icp and icm are not canceled. As described above, the problem of removing the first error is referred to as a first problem.

  Second, when a large current flows through the electromagnet coil 455, the nonlinearity of the magnetic characteristics of the electromagnet core 450 becomes significant, as will be described later. This is the main cause, and Np (icp) and Nm (icm), which are amounts of modulation of the P-side and M-side control current components by the sensor carrier, are included in the displacement signal of the rotating body as errors. This is the cause of the second error. As described above, the problem of removing the second error is referred to as a second problem.

From the above, when considering the first problem and the second problem, the expression (6) is transformed into the following expression (6a).
Ip = ib + icp-B {(Dd) + Np (icp)} × sin (ωc × t-π / 2)
Im = ib−icm + B {(D + d) + Nm (icm)} × sin (ωc × t−π / 2) (6a)

Since Ip and Im can be expressed as shown in equation (6a), a sum signal (Ip + Im) described later is expressed by equation (7) below.
Ip + Im = 2 × ib + △ icpm + B {2d-Np (icp) + Nm (icm)} × sin (ωc × t-π / 2)
... (7)
Here, Δicpm = icp−icm.

In Expression (7), Δicpm that causes the first error appears. In addition, [−Np (icp) + Nm (icm)] as the second error appears. The sum signal shown in Equation (7) is attenuated by Δicpm by passing through a filter such as a BPF centered on the sensor carrier frequency fc before demodulation. Therefore, Δicpm after attenuation is set to R (Δicpm). This is the first error.
From the above, the demodulated displacement signal is R (△ icpm) + B {2d-Np (icp) + Nm (icm)}, and R (△ icpm) + [-Np ( icp) + Nm (icm)].

  As a premise of the present invention, the first error R (Δicpm) is first removed using the invention described in Japanese Patent Application No. 2013-021681. The removal of the first error will be briefly described in the explanation of the AD converter 400 and the signal processing calculation unit 406 shown in FIG. Since details are described in Japanese Patent Application No. 2013-021681, detailed description is omitted here. The removal of [−Np (icp) + Nm (icm)], which is the second error, according to the present invention will be described later.

FIG. 8 is a functional block diagram of magnetic bearing control in the control unit 44 and shows one of the five control shafts. Here, magnetic bearing control during normal control will be described. The magnetic bearing control at the time of adjustment will be described later.
As described above, a pair of magnetic bearing electromagnets 45 on the P side and M side are provided for one control shaft, and an excitation amplifier 43 is provided for each magnetic bearing electromagnet 45. Current detection signals are output from the ten excitation amplifiers 43, respectively.

  The gate signal generation unit 401p of the control unit 44 generates a gate signal for driving the switching element of the P-side excitation amplifier 43. Similarly, the gate signal generation unit 401m generates a gate signal for driving the switching element of the M-side excitation amplifier 43. When the switching element of each excitation amplifier 43 is on / off controlled based on the gate signal, a voltage is applied to the electromagnet coil of the magnetic bearing electromagnet 45, and the above-described currents Ip and Im flow. From the current sensors 101A and 101B of the P-side excitation amplifier 43, a current detection signal (indicated by Ip similar to the current) of the current Ip flowing through the P-side magnetic bearing electromagnet 45 is output. On the other hand, from the current sensors 101A and 101B of the M-side excitation amplifier 43, a current detection signal (indicated by Im similar to the current) of the current Im flowing through the M-side magnetic bearing electromagnet 45 is output.

  The current detection signals Ip and Im output from the current sensors 101A and 101B are input to the AD converter 400 via low-pass filters 403 and 404, respectively. Further, the sum signal (Ip + Im) of the current detection signals Ip and Im that have passed through the low-pass filters 403 and 404 is input to the AD converter 400 via a band-pass filter 405 having the sensor carrier frequency fc as the center frequency. These signals Ip, Im, (Ip + Im) correspond to the electromagnet current signal 444 in FIG.

  When the current detection signals Ip and Im are passed through the low-pass filters 403 and 404, high-frequency noise components are removed. When the sum signal (Ip + Im) is passed through the bandpass filter 405, the bias component (2 × ib) contained in the sum signal (Ip + Im) is removed. Incidentally, 405 may be constituted by a high-pass filter instead of the band-pass filter in that the bias component is removed.

The AD converter 400 takes in the current detection signals Ip and Im and the sum signal (Ip + Im) at the sampling frequency fs. Here, the sampling frequency fs and the sensor carrier frequency fc are
fc = (n + 1/2) x fs (8)
There is a relationship. However, n is an integer greater than or equal to 0.
Based on the above equation (8), the AD converter 400 samples and captures the data value d1 near the maximum peak position and the data value d2 near the minimum peak position. The sum signal (Ip + Im) sampled by the AD converter 400 is input to the signal processing calculation unit 406. The signal processing calculation unit 406 calculates the displacement information of the shaft 5 based on the sampling data.

The signal processing calculation unit 406 calculates the following equation (9) as a demodulation calculation output based on the data value d1 in the vicinity of the maximum peak position and the data value d2 in the vicinity of the minimum peak position acquired by sampling of the AD converter 400. To do.
d3 = (d1-d2) / 2 (9)
However, when the output timing is the maximum peak timing, the data value d1 immediately before one sampling is applied, and when the output timing is the minimum peak timing, the data value d2 immediately before one sampling is applied. In the calculation shown in the above equation (9), the control current component lower than the sensor carrier is canceled even if the magnitude is asymmetric on the P side and M side. As a result, R (Δicpm), which is the first error, can be removed. This is as described in Japanese Patent Application No. 2013-021681.

  The current detection signals Ip and Im sampled by the AD converter 400 are input to the corresponding signal processing arithmetic units 409p and 409m, respectively. The signal processing calculation units 409p and 409m remove the sensor carrier component is based on the sampling data, and calculate and output only information related to the bias current ib and the flying control current ic that are current components contributing to the flying control force.

  The calculation result of the signal processing calculation unit 409p is subtracted from the output from the current limit circuit 408p by the difference unit 420p after passing through the amplifier controller 410p. Further, the subtracter 421p subtracts the sensor carrier component (v × sin (ωc × t)) from the sensor carrier generation circuit 411 from the subtraction processing result, and the PWM control signal is converted into a PWM operation unit based on the subtraction result. It is generated at 412p. The gate signal generation unit 401p generates a gate drive voltage, that is, a gate signal based on the PWM control signal generated by the PWM calculation unit 412p. An adder 422p is provided between the difference unit 421p and the PWM calculation unit 412p. However, the adder 422p has only a function of passing a signal except when performing excitation by an exciter 417p described later.

  In addition, the calculation result of the signal processing calculation unit 409m passes through the amplifier controller 410m, and is then subtracted from the output from the current limit circuit 408m by the difference unit 420m. Further, the adder 421m adds the sensor carrier component (v × sin (ωc × t)) from the sensor carrier generation circuit 411 to the subtraction processing result, and the PWM control signal is subjected to PWM calculation based on the addition result. It is generated in the part 412m. The gate signal generation unit 401m generates a gate drive voltage based on the PWM control signal generated by the PWM calculation unit 412m. Note that an adder 422m is provided between the adder 421m and the PWM calculation unit 412m. However, the adder 422m has only a function of passing a signal except when it is vibrated by a vibrator 417m described later.

  The calculation results of the signal processing calculation units 406, 409p, and 409m are input to the correction processing calculation unit 415, respectively. The correction processing calculation unit 415 generates a correction signal for removing the second error, using the signal Ip and the signal Im. Further, the correction processing calculation unit 415 performs calculation for subtracting the correction signal for removing the second error from the displacement signal obtained from the signal Ip + Im. As a result, the second error [−Np (icp) + Nm (icm)] included in the displacement signal obtained from the signal Ip + Im can be removed, and a true estimated displacement signal can be output. Note that the calculation performed by the correction processing calculation unit 415 will be described in detail in the explanation of the second error removal described later.

The levitation controller 407 generates the levitation control current setting by proportional control, integral control, differential control, or the like based on the true estimated displacement signal output from the correction processing calculation unit 415. For the P-side control, the difference unit 419p uses the bias current setting amount subtracted from the levitation control current setting. For the M-side control, the difference unit 419m sets the levitation control current setting to the bias current setting amount. The one added is used.
The switches 418p and 418m are normally on as shown in the figure. However, during the rough adjustment of the adjustment described later, it is turned off to block the flying control.

Further, when the pump unit 1 is disposed sideways or when the shaft 5 is greatly displaced due to excessive disturbance or the like applied to the pump, the exciting current on one of the opposing magnetic bearing electromagnets 45, for example, on the P side becomes large. Current limit circuits 408p and 408m are provided so that the M-side excitation current, which is the other, is zero and the sensor carrier component voltage is not applied to only one side. By providing the current limit circuits 408p and 408m, the sensor carrier voltage is always applied to the magnetic bearing electromagnet 45.

(Removal of second error)
Here, the removal of the second error will be described in the following order.
-Cause of second error-Elimination of second error by adjustment / correction signal

<Cause of second error>
FIG. 9A is a BH curve of the laminated silicon steel plate constituting the electromagnet core 450, and the slope of the tangent to this BH curve is the magnetic permeability μ. Here, μ = μr × μ0. μr represents relative permeability. μ0 is a constant value of magnetic permeability in vacuum. The BH curve of the laminated silicon steel sheet can be approximated by a straight line in a region where the magnetic field H is small, and has a large inclination. However, as the magnetic field H increases, the inclination decreases.

  FIG. 9B shows the dependence of the excitation current I flowing in the electromagnetic coil 450 with respect to 1 / μr, which is the inverse quantity of the relative permeability μr. Since the magnetic field H is approximately proportional to the excitation current I, the relationship between 1 / μr and the magnetic field H behaves in the same manner as in FIG. In FIG. 9B, a flat portion is seen in a region where the excitation current I is small. This corresponds to a region where the magnetic field H of the BH curve shown in FIG. In FIG. 9B, 1 / μr increases as the excitation current I increases. In summary, 1 / μr can be regarded as a constant when only the region where the excitation current I is small is considered, but 1 / μr cannot be regarded as a constant when the region where the excitation current I is large is considered. This is the cause of the second error.

<Adjustment>
By the way, the inverse quantities 1 / Lp and 1 / Lm of the inductance of the electromagnet coil 455 are obtained by changing the above equation (3) to the following equation (3a) when handling the region where the current I of the electromagnet coil 455 is large. To the non-linear magnetic characteristic of the relative permeability of the laminated silicon steel sheet.
1 / Lp = A × {(Dd) + lp / μrp}
1 / Lm = A × {(D + d) + lm / μrm} (3a)
Here, lp and lm are constants having dimensions of distances related to the P-side and M-side electromagnetic coils, respectively. μrp and μrm are the relative magnetic permeability of the P-side and M-side electromagnet cores, respectively.

  Here, the second error and the correction signal for removing the second error are related as shown below. As described above, in order to generate a correction signal for removing the second error, the above equation (3) is corrected to the above equation (3a).

As a result, the equations (5) and (6) are also corrected as the following equations (5a) and (6b).
isp = -v × sin (ωc × t-π / 2) / (ωc × Lp)
=-B {(Dd) + lp / μrp} × sin (ωc × t-π / 2)
ism = v × sin (ωc × t-π / 2) / (ωc × Lm)
= B {(D + d) + lm / μrm} × sin (ωc × t-π / 2) (5a)

Ip = ib + ic-B {(Dd) + lp / μrp} × sin (ωc × t-π / 2)
Im = ib-ic + B {(D + d) + lm / [mu] rm} * sin ([omega] c * t- [pi] / 2) (6b)

Therefore, the following formula (10) can be derived from the formula (6b).
Ip + Im = 2 × ib + △ icpm + B {2d-lp / μrp + lm / μrm} × sin (ωc × t-π / 2)
... (10)
Comparing equation (7) and equation (10),
[-Np (icp) + Nm (icm)] = (-lp / μrp + lm / μrm) (11)
It becomes. That is, it is understood that the second error [−Np (icp) + Nm (icm)] is caused by (−lp / μrp + lm / μrm).

In order to remove the second error, it is necessary to more accurately grasp the relationship between 1 / μr and current shown in FIG. That is, it is necessary to approximate the relationship between 1 / μr and current.
Here, in order to approximate the relationship between 1 / μr and current shown in FIG. 9B, 1 / μr is expanded to a power series. Specifically, as shown in the following equation (12), lp / μrp and lm / μrm are expanded in power series with currents Ip and Im.
lp / μrp = a0p + a1p × Ip + a2p × Ip ² + a3p × Ip ³ +…
lm / μrm = a0m + a1m × Im + a2m × Im ² + a3m × Im ³ + (12)
Since the expansion includes lp and lm, the coefficients a0p, a1p, a2p, a3p,..., a0m, a1m, a2m, a3m,. In the adjustment, optimization of these coefficients a0p, a1p, a2p, a3p,..., A0m, a1m, a2m, a3m,.

  The adjustment method will be described in detail with reference to FIG. 8, FIG. 10, and FIG. Description of the same parts as those in the normal control will be omitted in FIG. Adjustments are made by the manufacturer at the time of shipping inspection, or periodically by the user after delivery of the vacuum pump. The arrangement of the vacuum pump at the time of adjustment may be matched with the arrangement at the time of normal control, that is, the arrangement when using the vacuum pump.

  Adjustment is divided into coarse adjustment and fine adjustment described later. After coarse adjustment, fine adjustment is performed. In coarse adjustment, the above-described coefficients a0p, a1p, a2p, a3p,..., A0m, a1m, a2m, a3m,. In the fine adjustment, a correction signal obtained from a signal Ip described later, that is, a correction signal obtained from a P-side correction signal and a correction signal obtained from a signal Im described later, that is, an overall gain ratio r described later that controls the gain balance of the M-side correction signal is optimized. Turn into.

―Coarse tone―
FIG. 10 shows the position of the shaft 5 of the rotating body unit R during adjustment. Here, as an example, a state in which the axes of the electromagnets 451p and 451m facing the magnetic bearing 67 are adjusted is depicted. 10A and 10B show the state of coarse adjustment, and FIG. 10C shows the state of fine adjustment described later. The movable boundary 600 illustrated in FIGS. 10A to 10C is determined by the touch-down bearings 66a and 66b and is the movable boundary of the shaft 5. That is, the shaft 5 can move in the movable boundary 600.

  Here, the rough adjustment on the P side will be described first. In coarse adjustment, the switches 418p and 418m shown in FIG. Then, as shown in FIG. 10 (a), the shaft 5 is brought into contact with the touchdown bearings 66a and 66b so as to reach the movable boundary 600 on the P-side electromagnet 451p side which is the electromagnet on the adjustment side. Move. At this time, in order to maintain this attracting contact state, a large direct current flows through the electromagnet coil 455 of the P-side electromagnet 451p, which is the attracting-side electromagnet, according to a command from the bias current setting unit 423 shown in FIG. . On the other hand, the current value of the M-side electromagnet 451m which is the opposite side is set to the i_limit value by the bias current setting unit 423 shown in FIG. Incidentally, FIG. 7 shows the control signal dependency of the current flowing in the electromagnetic coil 455 in the current limit circuits 408p and 408m during the levitation control. As described above, during the rough adjustment in the non-levitation control state, the P side Since the bias current value is different on the M side, it is different from FIG. That is, at the time of coarse adjustment, the current limit circuits 408p and 408m operate so that signals of different values output from the bias current setting unit 423 to the P side and the M side are passed.

  Thus, the rotating body unit R can be kept stationary by maintaining the suction state. That is, the true displacement d can be prevented from changing with time. This suction state is maintained even if the vibration described later is applied.

  In the above-described suction contact state, the exciter 417p shown in FIG. 8 applies a signal having a predetermined frequency to the P-side electromagnet 451p on the adjustment side in the adder 422p and applies current fluctuation, that is, excitation. Apply. On the other hand, the adder 422m passes the signal without operating the vibrator 417m acting on the M-side excitation amplifier.

  The signals Ip and Im output from the excitation amplifier 43 pass through the AD converter 400 and the signal processing calculation units 409p and 409m and are input to the correction processing calculation unit 415, as in normal control.

On the other hand, the sum signal Ip + Im also passes through the band-pass filter 405, the AD converter 400, and the signal processing calculation unit 406 and is input to the correction processing calculation unit 415, as in normal control. This will be described in detail below.
The sum signal Ip + Im is given by Equation (7), that is,
Ip + Im = 2 × ib + △ icpm + B {2d-Np (icp) + Nm (icm)} × sin (ωc × t-π / 2)
... (7)
It is represented by

When the sum signal Ip + Im passes through the bandpass filter 405, the signal becomes R (Δicpm) + B {2d−Np (icp) + Nm (icm)}.
Further, by passing through the AD converter 400 and the signal processing operation unit 406, the first error R (Δicpm) is removed, and B {2d−Np (icp) + Nm (icm)}. That is, the second estimated error signal [−Np (icp) + Nm (icm)] is added to the true estimated displacement signal 2Bd.
This signal is input to the correction processing calculation unit 415.

Hereinafter, optimization in the correction processing calculation unit 415 will be described below.
As described above, when the signal resulting from the sum signal Ip + Im is input to the correction processing calculation unit 415, the second error [−Np (icp) + Nm (icm) is added to the true estimated displacement signal 2Bd. ] Is added. Since the true estimated displacement d does not change with time as described above, the true estimated displacement signal 2Bd has only a DC component. Therefore, it can be seen that the AC component of the signal caused by the sum signal Ip + Im is caused only by the second error [−Np (icp) + Nm (icm)]. The correction processing calculation unit 415 at the time of adjustment removes the true estimated displacement signal 2Bd that is a DC component and extracts a second error [−Np (icp) + Nm (icm)] that is an AC component.

In the adjustment, the correction processing calculation unit 415 performs optimization by comparing the second error obtained from the sum signal Ip + Im with the correction signal generated by the signals Ip and Im. The sum signal Ip + Im is affected by attenuation by passing through the band-pass filter 405 before being input to the correction processing calculation unit 415. On the other hand, the signals Ip and Im are not affected. Therefore, in order to compare the second error and the correction signal, the correction processing calculation unit 415 modifies Expression (11) as Expression (13) below.
[-Np (icp) + Nm (icm)] = Q (-lp / μrp + lm / μrm) (13)
here,
Q = b / (1 + Ts) (14)
It is.

The function Q is a transfer function of a low-pass filter for correcting the influence of attenuation when passing through the band-pass filter 405. b is called an overall gain, and is a variable at the time of coarse adjustment, but is a constant of a value determined by the coarse adjustment at the time of fine adjustment described later or during normal control. T is a time constant of the low-pass filter, which is a variable during coarse adjustment, but is a constant determined by coarse adjustment during fine adjustment or normal control. Further, s is a complex variable of Laplace transform, and s = jx2πxf where j is an imaginary number and the frequency of the applied signal is f. In addition, since the transfer function Q is usually incorporated as a digital filter, it is not explicitly expressed in the configuration of Expression (14), but is expressed explicitly only during adjustment for obtaining the coefficient.

Substituting equation (14) into equation (13) yields the following equation (15).
[-Np (icp) + Nm (icm)] = {b / (1 + Ts)} × (-lp / μrp + lm / μrm) (15)

On the other hand, the signals Ip and Im are expressed by Equation (12),
lp / μrp = a0p + a1p × Ip + a2p × Ip ² + a3p × Ip ³ +…
lm / μrm = a0m + a1m × Im + a2m × Im ² + a3m × Im ³ + (12)
As shown in FIG. 4, it is related to lp / μrp and lm / μrm.

Substituting equation (12) into equation (15) yields equation (16) below.
[-Np (icp) + Nm (icm)] = {b / (1 + Ts)} × {-(a0p + a1p × Ip + a2p × Ip ² + a3p × Ip ³ +…) + (a0m + a1m × ^{im + a2m × im 2 + a3m} × im 3 + ...)} ··· (16)

  FIG. 11 shows the operation of the correction processing calculation unit 415 during adjustment. In the following, first, the operation of the correction processing calculation unit 415 during coarse adjustment on the P side will be described with reference to FIG. As shown in FIG. 11, the correction processing calculation unit 415 includes signal conversion units 801 and 802, gain adjustment units 803 and 804, a difference unit 805, a correction unit 806, a high-pass filter calculation unit 807, and a difference unit 808. And a calculation unit 809.

  The signal Ip is input to the signal conversion unit 801, and the signal Ip is converted into the signal lp / μrp by the above equation (12) stored in the signal conversion unit 801 or a simplified equation thereof and output. The signal lp / μrp is input to the gain adjustment unit 803, but the gain adjustment unit 803 does not operate during coarse adjustment, and the signal lp / μrp is output as it is. As will be described later, this means that the overall gain ratio r is set to zero. In this specification, the signal lp / μrp is also referred to as a P-side correction signal.

  The signal Im is input to the signal conversion unit 802, and the signal Im is converted into an M-side correction signal lm / μrm by the above-described equation (12) stored in the signal conversion unit 802 or a simplified equation thereof, and is output. The signal lm / μrm is input to the gain adjustment unit 804, but the gain adjustment unit 804 does not operate during coarse adjustment, and the signal lm / μrm is output as it is. As will be described later, this means that the overall gain ratio r is set to zero. In this specification, the signal lm / μrm is also called an M-side correction signal.

  The signal lp / μrp and the signal lm / μrm are respectively input to the differencer 805, and the differencer 805 subtracts the signal lp / μrp from the signal lm / μrm, and the resulting signal (−lp / μrp + lm / μrm) Is output.

  The signal (−lp / μrp + lm / μrm) is input to the correction unit 806, and the low-pass filter transfer function Q stored in the correction unit 806, that is, the digital corresponding to the transfer function {b / (1 + Ts)} A correction signal Q (−lp / μrp + lm / μrm) is output as a result of correction by the filter operation. Furthermore, the correction signal Q (−lp / μrp + lm / μrm) is removed by passing a direct current or extremely low frequency (for example, 10 Hz or less) component through the high-pass filter calculator 807 and is input to the differencer 808. The

On the other hand, the signal resulting from the sum signal Ip + Im is input to the calculation unit 809, where the true estimated displacement signal 2Bd, which is a DC component, is removed by the calculation unit 809, and the second error shown on the left side of the above equation (15). Only [-Np (icp) + Nm (icm)] is output. The second error [−Np (icp) + Nm (icm)] is input to the differentiator 808.

  In the differentiator 808, the right side is subtracted from the left side shown in Expression (13). That is, the correction signal Q (−lp / μrp + lm / μrm) on the right side is subtracted from the second error [−Np (icp) + Nm (icm)] on the left side.

As described above, here, the coarse adjustment on the P side, that is, the adjustment of the coefficients a0p, a1p, a2p, a3p,..., Which are the P side coefficients, that is, the P side coefficient adjustment is performed. In the case of coarse adjustment on the P side, Ip is a large current several times as large as ib, while Im is i_limit and is a small current. Therefore, (a0m +
a1m × Im + a2m × Im ² +…) is small. For this reason, the coefficients a0m, a1m, a2m, a3m,... Are initially fixed as, for example, zero, and the optimization calculation for the coefficients a0p, a1p, a2p, a3p,. Optimization processing by the square method can be performed.

  Specifically, bias currents ib of various magnitudes are set from the bias current setting unit 423 in FIG. Further, an excitation signal is applied from the vibrator 417p of FIG. 8 while changing in various sizes, and is input to Ip of FIG. 11, and is stored in the signal conversion unit 801 shown in FIG. 11 (12) The coefficients a0p, a1p, a2p, a3p,. The above-described calculation shown in FIG. 11 is repeated to obtain optimum coefficients a0p, a1p, a2p, a3p,... For reducing the subtraction result to a predetermined predetermined allowable level. The optimized coefficients a0p, a1p, a2p, a3p,... Are stored in the signal conversion unit 801.

  When the coarse adjustment on the P side is completed, as shown in FIG. 10B, the M-side electromagnet 451m on the opposite side attracts the shaft 5 of the rotating body unit R, so that the shaft 5 can be moved to the electromagnet 451m side. Move to boundary 600. Then, the vibrator 417m shown in FIG. At that time, the vibrator 417p is not operated. After that, the coarse adjustment on the M side, that is, the adjustment of the coefficients a0m, a1m, a2m, a3m,.

  Similarly, in the coarse adjustment on the M side, the bias current ib of various magnitudes is set from the bias current setting unit 423 in FIG. 8 within a range in which the suction state can be maintained. Further, an excitation signal is applied from the vibrator 417m shown in FIG. 8 while being changed in various sizes, and is input to Im shown in FIG. 11, and the equation (12) stored in the signal conversion unit 802 shown in FIG. The coefficients a0m, a1m, a2m, a3m, etc. are calculated and calculated by the method of least squares. The above-described calculation shown in FIG. 11 is repeated to obtain optimum coefficients a0m, a1m, a2m, a3m,... For reducing the subtraction result of the differentiator 808 to a predetermined allowable level. The optimized coefficients a0m, a1m, a2m, a3m,... Are stored in the signal conversion unit 802. Needless to say, the coefficients a0p, a1p, a2p, a3p,... Optimized in the coarse adjustment on the P side are stored in the signal calculation unit 801 during the coarse adjustment on the M side.

  When the coefficients a0p, a1p, a2p, a3p, ..., a0m, a1m, a2m, a3m, ... are optimized by the coarse adjustment on the P side and M side, the subtraction result of the differencer 808 is further reduced. The transfer function Q of the low-pass filter stored in the unit 806, that is, the overall gain b and the time constant T of the low-pass filter are optimized, and the values are newly stored. At this time, the optimized overall gain b is also stored in the gain adjusters 803 and 804.

The coarse adjustment shown above is performed from the P-side electromagnet 451p, but may be performed from the M-side electromagnet 451m. Further adjustments can be made by lowering the allowable level in the process of repeating the same operation a plurality of times. Further, as shown in FIG. 9B, the inverse quantity of the relative permeability changes depending on the excitation current I. Therefore, it is better to provide a larger bias current for coarse adjustment. By storing coefficients optimized with various bias current setting values in each part of the correction processing calculation unit 415, it becomes easier to deal with various bias currents applied during normal control, and the second error is reduced. It becomes easy.

  When the coarse adjustment of the opposing electromagnets 451p and 451m of the magnetic bearing 67 is completed, the coarse adjustment is similarly performed on the opposing electromagnets 452p and 452m of the magnetic bearing 67 and the electromagnets 45 of the other magnetic bearings 68 and 69. However, the magnetic bearing 69 is similarly adjusted by moving the rotor disk 55 instead of the shaft 5.

―Fine adjustment―
Hereinafter, the fine adjustment performed after the rough adjustment will be described. After each coefficient is obtained in the above-described coarse adjustment, the switches 418p and 418m shown in FIG. 8 are turned on to switch to the levitation control, and as shown in FIG. 10C, the rotating body unit R including the shaft 5 is changed. In the state where it is levitated to a predetermined position, signals of a predetermined frequency with the same sign are added and superimposed on the P side and M side using the vibrators 417p and 417m shown in FIG. In this case, since the attractive force to be attracted to the P side and the attractive force to be attracted to the M side are equally balanced, the signal component for adjustment superimposed and added is small.

  However, in order to further adjust the gain balance on the P side and M side, the overall gain ratio r described later is optimized by fine adjustment. The overall gain ratio r is a variable during fine adjustment, but is set to zero during coarse adjustment, and is set to a constant value determined by fine adjustment during normal control.

  The fine adjustment will be specifically described with reference to FIG. A description of the same parts as the coarse adjustment is omitted. During fine adjustment, the signal gain adjustment unit 803 multiplies the signal lp / μrp by the coefficient (1 + r / b), and the signal gain adjustment unit 804 multiplies the signal lm / μrm by the coefficient (1-r / b). Is a big difference. Strictly speaking, although the above-described coefficient was multiplied even in the rough gradation, the overall gain ratio r was set to zero, so that it was easy to understand. As a result, the correction signal input to the differentiator 808 is an AC component of Q {− (1 + r / b) × lp / μrp + (1−r / b) × lm / μrm}. The gain adjusters 803 and 804 store not only the overall gain ratio r but also the overall gain b, but the overall gain b optimized by coarse adjustment as described above is stored.

Therefore, as described above, equation (16) is modified into equation (17) shown below and used.
[-Np (icp) + Nm (icm)] = {b / (1 + Ts)} × {-(1 + r / b) × (a0p + a1p × Ip + a2p × Ip ² + a3p × Ip ³ + …) + (1-r / b) × (a0m + a1m × Im + a2m × Im ² + a3m × Im ³ +…)} (17)

  In the differentiator 808, the right side is subtracted from the left side shown in the above equation (17). That is, from the second error [−Np (icp) + Nm (icm)] on the left side, the correction signal Q {− (1 + r / b) × lp / μrp + (1-r / b) × lm / on the right side μrm} AC component is subtracted. Stored in gain adjusting sections 803 and 804 so that the subtraction result is reduced to a predetermined predetermined allowable level, that is, the amplitude value of the frequency component of the signal applied for excitation is equal to or lower than the predetermined level. By changing only the overall gain ratio r, the optimum overall gain ratio r is obtained so that the subtraction result of the differentiator 808 is below the allowable level.

That is, at the time of coarse adjustment, the coefficient a0p + a1p × Ip + a2p × Ip ² +... Stored in the signal conversion unit 801 and the coefficient a0m + a1m × Im + a2m × Im stored in the signal conversion unit 802. ² +... The coefficients of the low-pass filter transfer function Q (overall gain b, time constant T) stored in the correction unit 806 are optimized to change the gain. Adjustment is performed by changing the overall gain ratio r stored in the sections 803 and 804.

  When performing the fine adjustment, the overall gain ratio r is changed within the range of the overall gain ± b. When the overall gain ratio r is optimized, the optimum values are stored in the gain adjusters 803 and 804, respectively.

  After the fine adjustment of the opposing electromagnets 451p and 451m of the magnetic bearing 67 is completed, the fine adjustment is similarly performed on the opposing electromagnets 452p and 452m of the magnetic bearing 67 and the electromagnets 45 of the magnetic bearings 68 and 69. Further, as in the case of the coarse adjustment, it is preferable to carry out a plurality of times in order to improve the accuracy of the overall gain ratio r. Further, fine adjustment with various bias currents can accurately cope with the removal of the second error under various bias currents during normal control.

  As described above, the correction processing calculation unit 415 performs coefficients a0p, a1p, a2p, a3p,..., A0m, a1m, a2m, a3m,. The values of gain b, time constant T) and overall gain ratio r are stored. Then, as will be described later, the correction processing calculation unit 415 stores the coefficients a0p, a1p, a2p, a3p,..., A0m, a1m, a2m, a3m,. Using the gain b, time constant T), and overall gain ratio r, a true estimated displacement signal 2Bd from which the second error has been removed during normal control is generated and output (see FIG. 12).

―Specific examples of adjustment―
As described above, adjustment has been described, but a specific example at the time of adjustment will be described below.
As a simple example of the expression (12), the following expression (12a) is used.
lp / μrp = a0p + a1p × Ip + a2p × Ip ²
lm / μrm = a0m + a1m × Im + a2m × Im ² (12a)
further,
Q = b / (1 + Ts) (14)
Is used.

Then, from equation (16),
Correction signal = {b / (1 + Ts)} × {-(a0p + a1p × Ip + a2p × Ip ² ) + (a0m + a1m × Im + a2m × Im ² )} (18)
It becomes. Appropriate initial values are given to a0p, a1p, a2p, a0m, a1m, a2m, b, and T shown in the equation (18) to start adjustment.

  The levitation control is interrupted by turning off the switches 418p and 418m shown in FIG. Further, a DC current, for example, four times the bias current is supplied to the P side, and the shaft 5 of the rotating body unit R is attracted to the P side as shown in FIG. An excitation signal of 100 Hz, which is an initial excitation signal, is applied to the electromagnet 45 on the P side using the vibrator 417p shown in FIG. At this time, the current flowing through the M-side electromagnet coil is i_limit. In this state, the correction signal calculated by the above equation (18) from the second error extracted by excluding the true estimated displacement signal 2Bd that becomes only the DC component at the time of adjustment from the demodulated signal is converted to the high-pass filter calculator 807. A0p, a1p, and a2p are calculated by applying a least square method or the like on the data obtained by subtracting the AC component extracted by passing, that is, the difference between the second error and the correction signal.

  Similarly, as shown in FIG. 10B, the shaft 5 of the rotator unit R is moved to the M side and sucked to calculate a0m, a1m, and a2m. Depending on the individual magnetic bearing device, if necessary, in order to improve the accuracy of a0p, a1p, a2p, a0m, a1m, a2m, various bias current values are set within the range where the attraction state can be maintained. The above adjustment operation is repeated a plurality of times while changing to excitation signals having various amplitude values.

Next, the frequency of the excitation signal is increased and the above adjustment is repeated at, for example, 200 Hz. Coefficients a0p, a1p, a2p, a0m, a1m, and a2m are obtained under the condition that the frequency is sequentially increased to 1000 Hz. Coefficients a0p, a1p, a2p, a0m, a1m, and a2m are obtained at the respective excitation frequencies. The above optimization calculation is performed with various overall gains b and time constants T so that they match as much as possible. That is, the optimization calculation of the overall gain b and the time constant T is also performed. After the optimization calculation, the coefficients a0p, a1p, a2p, a0m, a1m, a2m, the overall gain b, and the time constant T are fixed.

Fine adjustment is performed after giving coefficients a0p, a1p, a2p, a0m, a1m, and a2m optimized by coarse adjustment. The switches 418p and 418m shown in FIG. 8 are turned on to float the rotator unit R, and the rotator unit R is brought to a static levitation state, and an excitation signal whose phase is matched to both the P side and the M side is sent. Application is performed using the vibrators 417p and 417m. The overall gain value ratio r of the correction signal shown in the equation (19) is optimized so that the excitation signal frequency component included in the signal after subtraction by the correction signal shown in the following equation (19) is below the allowable level. Turn into.
Correction signal = (b / (1 + Ts)} × {-(1 + r / b) × (a0p + a1p × Ip + a2p × Ip ² ) + (1-r / b) × (a0m + a1m × Im ^{+ a2m × Im 2)} ···} (19)
For example, the overall gain value ratio r is adjusted within a range of ± 20% of the overall gain b.

<Removal of second error by correction signal>
FIG. 12 shows the operation of the correction processing calculation unit 415 during normal control. The operation during normal control of the correction processing calculation unit 415 shown in FIG. 8 will be described with reference to FIG. The operation during normal control of the correction processing calculation unit 415 generates a correction signal for removing the second error using each parameter optimized by the above-described adjustment and the signals Ip and Im, and generates the correction signal from the signal Ip + Im. By subtracting the correction signal from the obtained displacement signal, the second error is removed and a true estimated displacement signal 2Bd is output. Details will be described below.

  As shown in FIG. 12, the correction processing calculation unit 415 includes signal conversion units 801 and 802, gain adjustment units 803 and 804, a difference unit 805, a correction unit 806, a high-pass filter calculation unit 807, and a difference unit 808. And a calculation unit 809. However, the calculation unit 809 is not illustrated because it does not operate during normal control.

  The signal Ip is input to the signal conversion unit 801, and the signal Ip is converted into the signal lp / μrp by the above equation (12) stored in the signal conversion unit 801 or a simplified equation thereof and output. The signal lp / μrp is input to the gain adjustment unit 803, multiplied by the coefficient (1 + r / b) stored in the gain adjustment unit 803, and output as a signal (1 + r / b) × lp / μrp. Is output.

  The signal Im is input to the signal conversion unit 802, and the signal Im is converted into the signal lm / μrm by the above-described equation (12) stored in the signal conversion unit 802 or a simplified equation thereof, and is output. The signal lm / μrm is input to the gain adjustment unit 804, multiplied by the coefficient (1-r / b) stored in the gain adjustment unit 804 and output, and the signal (1-r / b) × lm / μrm is Is output.

  The signal (1 + r / b) × lp / μrp and the signal (1-r / b) × lm / μrm are respectively input to the differentiator 805, and the differencer 805 outputs the signal (1-r / b) × lm / μrm. The signal (1 + r / b) x lp / μrp is subtracted from the resulting signal {-(1 + r / b) x lp / μrp + (1-r / b) x lm / μrm} Is done.

  The signal {− (1 + r / b) × lp / μrp + (1-r / b) × lm / μrm} is input to the correction unit 806 and stored in the correction unit 806, that is, the transfer function Q of the low-pass filter, that is, , Corrected by a digital filter operation corresponding to the transfer function {b / (1 + Ts)}, and the resulting correction signal Q {-(1 + r / b) × lp / μrp + (1-r / b ) × lm / μrm} is output. The correction signal Q {− (1 + r / b) × lp / μrp + (1−r / b) × lm / μrm} passes through the high-pass filter calculator 807, and an AC component is extracted to the differencer 808. Entered.

By optimizing various parameters by the adjustment described above, the second error [−Np (icp) + Nm (icm)] and the correction signal Q {− (1 + r / b) × lp / μrp + (1-r / b) × lm / μrm} holds the relationship shown in the following equation (20).
[−Np (icp) + Nm (icm)] = Q {− (1 + r / b) × lp / μrp + (1-r / b) × lm / μrm} (20)
Note that this is equivalent to equation (17), as can be seen from equations (12) and (14).

  On the other hand, a signal obtained from the sum signal Ip + Im is input to the differentiator 808. Note that when the signal obtained from the signal Ip + Im is input to the correction processing calculation unit 415, the first error R (Δicpm) has already been removed, and the signal obtained from the sum signal Ip + Im has a true value. The estimated displacement signal 2Bd is a signal obtained by adding a second error [-Np (icp) + Nm (icm)].

  In the differentiator 808, the correction signal Q {− (1 + r / b) × lp / μrp + (1-r / b) × lm / μrm} is subtracted from the signal obtained from the sum signal Ip + Im, that is, The second error is removed, and a true estimated displacement signal 2Bd is output.

  After the above-described adjustment is completed, the movement distance of the contact position, the gain of the signal corresponding to the floating center position, and the offset may be adjusted (the functional block is omitted). Also, gain offset adjustment can be performed from the DC level of the demodulated signal during the suction process.

  In the above-described adjustment, the correction signal approximation and accuracy improvement and the adjustment time have a trade-off relationship. If the adjustment time is loose, higher order approximation may be applied to improve accuracy (see Equation (12) and Equation (12a)).

  In the above embodiment, the present invention is mainly used for removing the second error. However, when removing the second error in the present invention, the first error can be removed at the same time. This is because, in the present embodiment, the correction signal is obtained by the optimization calculation in a state where only the second error is included in the true estimated displacement signal 2Bd, but the first error and the second error are estimated to be true. This is because the correction signal may be obtained by optimization calculation in a state where it is included in the displacement signal 2Bd. Therefore, the first error and the second error can be removed only by the present invention without using the invention described in Japanese Patent Application No. 2013-021681. However, as shown in the above embodiment, after removing the first error using the invention described in Japanese Patent Application No. 2013-021681, removing the second error using the present invention improves the accuracy. Needless to say, this is preferable. For example, although the invention described in Japanese Patent Application No. 2013-021681 is used, even if the first error is not sufficiently removed as a result for some reason, by implementing the present invention, only the second error can be obtained. The first error that could not be sufficiently removed can also be removed. As described above, the first error can be removed by the two inventions, and the accuracy can be further improved.

The adjustment in the present invention may be performed semi-automatically or semi-manually at the time of shipping inspection. Moreover, you may adjust regularly and automatically between the time of the use after shipment.
Further, in some cases, only the coarse adjustment can be performed and the fine adjustment can be omitted.

As described above, according to the magnetic bearing device of the control unit and the turbo molecular pump including the magnetic bearing device according to the present invention, the following operational effects can be achieved.
(1) The magnetic bearing device of the control unit according to the present invention is provided on each of the rotating body unit R and a plurality of control shafts for controlling the flying of the rotating body unit R, and each of the control shafts via the rotating body. PWM control is performed on the voltages applied to the P-side electromagnet 45 and the M-side electromagnet 45, and the P-side electromagnet 45 and the M-side electromagnet 45, which are arranged to face each other, and the floating position of the rotating body unit R is detected. The excitation amplifier 43 provided in each of the P-side electromagnet 45 and the M-side electromagnet 45 for supplying the electromagnet currents Ip and Im on which the sensor carrier signal for superimposition is provided, and the excitation amplifier 43 are provided respectively. Current sensors 101A and 101B that detect Ip and Im, and a true estimated displacement signal 2Bd based on the electromagnet currents Ip and Im, and an excitation amplifier 4 based on the true estimated displacement signal 2Bd The a control unit 44 for PWM control, a.
The AD converter 400 of the control unit 44 generates a detection signal Ip of the electromagnet current Ip flowing through the P-side electromagnet 45, a detection signal Im of the electromagnet current Im flowing through the M-side electromagnet 45, and the current signal Ip and the current signal Im. AD sampling is performed on the sum signal Ip + Im obtained by addition. At that time, the AD converter 400 AD samples the data value d1 near the maximum peak value and the data value d2 near the minimum peak value from the sum signal Ip + Im.
The signal processing calculation unit 406 of the control unit 44 removes the first error by obtaining the data value d3 by performing the calculation process of d3 = (d1-d2) / 2 using the data values d1 and d2. Further, the signal processing calculation unit 40 generates a demodulated displacement signal 2Bd + [− Np (icp) + Nm (icm)] based on the data value d3.
In the correction processing calculation unit 415 of the control unit 44, the signal conversion unit 801 generates the correction signal lp / μrp by calculation processing from the signal Ip after AD sampling. The signal conversion unit 802 generates a correction signal lm / μrm by arithmetic processing from the signal Im after AD sampling. The gain adjusting unit 803 multiplies the correction signal lp / μrp by (1 + r / b) to generate a signal (1 + r / b) × lp / μrp. The gain adjusting unit 804 multiplies the correction signal lm / μrm by (1-r / b) to generate a signal (1-r / b) × lm / μrm. The subtractor 805 subtracts the signal (1 + r / b) × lp / μrp from the signal (1-r / b) × lm / μrm to obtain a signal {− (1 + r / b) × lp / μrp + (1 -r / b) × lm / μrm}. The correction unit 806 calculates the signal {-(1 + r / b) × lp / μrp + (1-r / b) × lm / μrm} through the transfer function Q of the low-pass filter and corrects the correction signal Q {− (1 + r / b) × lp / μrp + (1-r / b) × lm / μrm}. Further, an AC component is extracted after passing through the high-pass filter calculator 807, and the differencer 808 extracts the correction signal Q {− (1 + r /) from the demodulated displacement signal 2Bd + [− Np (icp) + Nm (icm)]. b) × lp / μrp + (1-r / b) × lm / μrm} is subtracted to remove the second error [−Np (icp) + Nm (icm)], and the true estimated displacement signal 2Bd Get.
As described above, the AD converter 400 and the signal processing calculation unit 406 remove the first error, and the correction processing calculation unit 415 removes the second error, so that only one of the opposing electromagnet coils is large. Even when a current is passed, the first error and the second error can be accurately removed, and a good S / N ratio can be maintained with respect to the displacement signal of the rotating body.

(2) The rotating body unit R has a movable boundary 600. The correction signal lp / μrp is composed of a function (for example, the first expression of Expression (12)) including the signal Ip. The correction signal lm / μrm is composed of a function (for example, the second expression of Expression (12)) including the signal Im. When coarse adjustment on the P side where the coefficient used for the function consisting of the signal Ip is optimized is performed, the rotating body unit R is moved to the movable boundary 600 on the P-side electromagnet 45 side, and a predetermined bias current value is obtained. After the rotating body unit R is magnetically attracted and fixed by ib, coarse adjustment on the P side is performed. When the coarse adjustment on the M side in which the coefficient used for the function composed of the signal Im is optimized, the rotating body unit R is moved to the movable boundary 600 on the M electromagnet 45 side, and a predetermined bias current value is obtained. After the rotating body unit R is magnetically attracted and fixed by ib, coarse adjustment on the M side is performed.
By performing the coarse adjustment on the P side and the coarse adjustment on the M side, it is possible to appropriately remove the second error that occurs during the normal control using the predetermined bias current ib.

(3) The coarse adjustment on the P side is obtained by superimposing an applied signal having a predetermined frequency on the electromagnet current Ip on the P-side electromagnet 45 side and then applying the true estimated displacement signal 2Bd that becomes the applied frequency component. It is carried out so that the amplitude value is below an allowable level. In the coarse adjustment on the M side, an applied signal having a predetermined frequency is superimposed on the electromagnet current Im on the M-side electromagnet 45 side, and the amplitude value of the true estimated displacement signal 2Bd that becomes the applied frequency component is It is carried out so as to be below the allowable level.
As described above, the coefficient used for the function composed of the signal Ip, the coefficient used for the function composed of the signal Im, and the like can be easily optimized using the AC component.

(4) It is preferable that the coarse adjustment and the fine adjustment are performed with a plurality of bias current values ib. Thereby, the non-linearity of the magnetic characteristics of the electromagnet core can be grasped more accurately, and the second error can be more accurately removed at various bias current values ib during normal control. It should be noted that the more the number of bias current values ib to be implemented is, the more accurately the nonlinearity of the magnetic characteristics of the electromagnet core described above can be grasped.

(5) The coarse adjustment and the fine adjustment may be performed a plurality of times, that is, repeatedly at a predetermined bias current value ib. As a result, the bias current value ib can be optimized more accurately, and the second error during normal control can be more accurately removed.

(6) The magnetic bearing device of the control unit of the present invention performs magnetic levitation control so that the position of the rotary unit R does not change in time after the P-side coarse adjustment and the M-side coarse adjustment are completed. Further, an applied signal having a predetermined frequency is superimposed on each of the electromagnet currents Ip and Im flowing in the P-side electromagnet 45 and the M-side electromagnet 45. At that time, the magnetic attractive force to the P side applied to the rotator unit R caused by the above-described applied signal superimposed on the current Ip and the rotator unit R caused by the above-described applied signal superimposed on the current Im. The magnetic attraction forces toward the M side are set so as to cancel each other, and as a result, the magnetic levitation control is performed so that the position of the rotating body unit R is not changed with time. After that, the gain balance between the correction signal lp / μrp and the correction signal lm / μrm in the correction signal Q {-(1 + r / b) × lp / μrp + (1-r / b) × lm / μrm} A fine adjustment for adjusting the overall gain ratio r is performed.
By this fine adjustment, the correction signal Q {-(1 + r / b) × lp / μrp + (1-r / b) × lm / μrm} is optimal for removing the second error. The second error can be removed more accurately.

(7) Since the turbo molecular pump includes the magnetic bearing device of the control unit according to the present invention, the first error and the second error can be accurately detected even when a large current is supplied only to one of the opposing electromagnetic coils. And a good S / N ratio can be maintained for the displacement signal of the rotating body. Furthermore, the vibration of the pump casing can be reduced, and the effect of improving the floating control performance of the rotating body can be achieved.

  In addition, the above description is an example to the last, and this invention is not limited to the said embodiment at all unless the characteristic of this invention is impaired. For example, in the above-described embodiment, the turbo molecular pump having the turbo pump unit and the drag pump unit has been described as an example. it can.

1: pump unit, 4: pump rotor, 4a: rotor blade, 4b: cylindrical part,
5: Shaft, 40: DC power supply, 41: Inverter, 42: Motor,
42a: motor stator, 42b: motor rotor, 43: excitation amplifier,
44: control unit, 45: magnetic bearing electromagnet, 55: rotor disk, 60: base,
60a: exhaust port, 61: pump casing, 61a: intake port, 61b: locking part,
61c: fixed flange, 62: fixed wing, 63: spacer ring,
64: Screw stator, 65: Exhaust port,
66a, 66b: touchdown bearings, 67, 68, 69: magnetic bearings,
101A, 101B: current sensor, 400: AD converter,
401p, 401m: gate signal generation unit, 403, 404: low-pass filter,
405: Band pass filter, 406: Signal processing operation unit, 407: Levitation controller,
408p, 408m: current limit circuit, 409p, 409m: signal processing operation unit,
410p, 410m: amplifier controller, 415: correction processing calculation unit,
417p, 417m: vibrator, 418p, 418m: switch,
420p, 420m, 421p: differentiator,
421m, 422p, 422m: adder, 423: bias current setting unit,
441: PWM control signal, 442: signal, PWM gate drive signal 443,
444: electromagnet current signal, 450: electromagnet core,
451p, 451m, 452p, 452m: magnetic bearing electromagnet, 455: electromagnet coil, 600: movable boundary, 801, 802: signal conversion unit, 803, 804: gain adjustment unit, 805: difference unit, 806: correction unit, 807: High-pass filter calculator,
808: differentiator, 809: arithmetic unit,
d: displacement, D10, D11: diode, J: levitation center axis (levitation target position),
L1, L2: Line, R: Rotating body unit, SW10, SW11: Switching element

Claims (8)

A rotating body,
A first electromagnet and a second electromagnet provided on each of a plurality of control shafts for controlling the levitation of the rotator, and arranged to face each of the control shafts via the rotator;
The first electromagnet and the second electromagnet that supply the electromagnet current superimposed with the sensor carrier signal for detecting the floating position of the rotating body by PWM controlling the voltage applied to each of the first electromagnet and the second electromagnet. An excitation amplifier provided in each of the two electromagnets;
A current sensor provided in each of the excitation amplifiers to detect the electromagnet current;
A controller that obtains an estimated displacement signal based on the electromagnet current and performs PWM control of the excitation amplifier based on the estimated displacement signal ;
The controller is
In obtaining the estimated displacement signal in each control axis of the plurality of control axes ,
A detection signal of the electromagnet current flowing through the first electromagnet (hereinafter referred to as a first current signal), a detection signal of the electromagnet current flowing through the second electromagnet (hereinafter referred to as a second current signal), and the first current signal; AD sampling is performed in synchronization with the sensor carrier signal for the sum signal obtained by adding the second current signal,
A demodulated displacement signal is generated by a sum signal AD-synchronized with the sensor carrier signal ;
The first correction signal for reducing the influence on the demodulated displacement signal due to the non-linearity of the magnetic characteristics of the first electromagnet is the first current signal after the AD sampling and the magnetic characteristics of the first electromagnet. Generated by arithmetic processing based on nonlinearity ,
The second correction signal for reducing the influence on the demodulated displacement signal due to the non-linearity of the magnetic characteristics of the second electromagnet is the second current signal after the AD sampling and the magnetic characteristics of the second electromagnet. Generated by arithmetic processing based on nonlinearity ,
A third correction signal as a difference between the first correction signal and the second correction signal is generated by arithmetic processing;
By subtracting the third correction signal from the demodulated displacement signal, a magnetic bearing device for obtaining the estimated displacement signals. The magnetic bearing device according to claim 1,
The sum signal has a sampling frequency fs satisfying fc = (n + 1/2) · fs (where n is an integer of 0 or more), where the frequency of the sensor carrier signal is fc, and the sensor AD sampling is performed at a timing near the maximum peak position and a timing near the minimum peak position of the carrier signal,
When the data value of the sum signal sampled in the vicinity of the maximum peak position is d1, and the data value of the sum signal sampled in the vicinity of the minimum peak position is d2, it is calculated by d3 = (d1−d2) / 2. A magnetic bearing device in which the value d3 is the demodulated displacement signal . The magnetic bearing device according to claim 1 or 2,
The rotating body has a movable boundary;
The first correction signal is composed of a function (hereinafter referred to as a first function) composed of the first current signal.
The second correction signal is composed of a function (hereinafter referred to as a second function) composed of the second current signal,
When an adjustment that optimizes a coefficient used for the first function (hereinafter referred to as a first coefficient adjustment) is performed, the rotating body is moved to the movable boundary of the rotating body on the first electromagnet side, The first coefficient adjustment is performed after magnetically attracting and fixing the rotating body with a predetermined bias current value,
When an adjustment that optimizes a coefficient used for the second function (hereinafter referred to as a second coefficient adjustment) is performed, the rotating body is moved to the movable boundary of the rotating body on the second electromagnet side, A magnetic bearing device in which the second coefficient adjustment is performed after the rotating body is magnetically attracted and fixed at the predetermined bias current value. The magnetic bearing device according to claim 3,
In the first coefficient adjustment, an applied signal having a predetermined frequency is superimposed on at least the electromagnet current on the first electromagnet side, and the amplitude of the estimated displacement signal having the predetermined frequency is below a predetermined level. Carried out to be
In the second coefficient adjustment, an applied signal having the predetermined frequency is superimposed on at least the electromagnet current on the second electromagnet side, and the amplitude of the estimated displacement signal having the predetermined frequency is below a predetermined level. Magnetic bearing device implemented to become. The magnetic bearing device according to claim 4,
A plurality of the predetermined bias current values are provided,
The first coefficient adjustment and the second coefficient adjustment are performed at each of the plurality of predetermined bias current values provided. The magnetic bearing device according to claim 4 or 5 ,
The magnetic bearing device is configured such that the first coefficient adjustment and the second coefficient adjustment are performed a plurality of times at the predetermined bias current value. In the magnetic bearing device according to any one of claims 4 to 6,
The third correction signal is generated by calculating a difference between the first correction signal and the second correction signal at a predetermined ratio,
The predetermined ratio is a magnetic attraction when the applied signal having the predetermined frequency is superimposed on the electromagnet current flowing in the first electromagnet in a state where the magnetic levitation control is performed so that the position of the rotating body does not change with time. force and a magnetic attraction force in the case of superimposing the applied signal having the predetermined frequency to the electromagnet current flowing through the second electromagnet Ru is set to be canceled, the magnetic bearing device. A pump rotor formed with an exhaust function part;
A motor for rotationally driving the pump rotor;
A vacuum pump comprising: the magnetic bearing device according to claim 1, wherein the magnetic rotor device supports the rotor of the pump rotor in a magnetically levitated manner.

Images