TWI735223B - Motor control device, notch filter adjustment device, and notch filter adjustment method - Google Patents

Abstract

A motor control device includes: a controller that controls a control object including a motor; and a vibration extraction unit that extracts one or more vibrations caused by the resonance characteristics of one or more controlled objects and superimposed on the response of the control system Component; the successive frequency estimation unit, which is in the vibration component, successively estimates the frequency of a certain vibration component, and uses its output as the vibration frequency estimation value sequence; and the resonance number estimation unit, which is based on the vibration frequency estimation value sequence, The number of resonance characteristics that become the cause of vibration that overlaps the response of the control system is output as the resonance number estimation value sequence, and the number of notch filters corresponding to the value of the resonance number estimation value sequence is set; the controller's The output is given to the current controller through the notch filter to control the motor.

Description

Motor control device, notch filter adjustment device, and notch filter adjustment method

The present invention relates to motor control.

In recent years, in terms of FA, it is expected to shorten the introduction time of the motor control system and increase the productivity due to the shortened production interval time caused by the optimal adjustment of the motor control system. One of the adjustment requirements of the motor control system is the parameter of the control means that suppresses the resonance of the mechanical system. A technology that does not artificially adjust automatically in a short time and optimally becomes a solution to the above-mentioned demand.

Generally, it is because the resonance characteristics of the mechanical system cannot increase the gain of the feedback controller (hereinafter, abbreviated as FB controller). In order to avoid this situation, the notch filter is placed in the FB control. The rear section of the device to offset the resonance characteristics. However, the filter parameters of the notch filter must be appropriately set for the resonance characteristics.

In addition, there may be multiple resonance characteristics of the mechanical system, and it is necessary to apply a notch filter to all of the obstacles to the increase in the gain of the FB controller.

Therefore, the automatic adjustment of the control means to suppress the resonance of the mechanical system mentioned above is based on the optimization of the filter parameters of the notch filters between the number of notch filters in the back stage of the FB controller and the intervening notch filters. constitute.

As a means of performing such automatic adjustment, patent documents 1 and 2 have been proposed. [Prior Technical Literature] [Patent Literature]

[Patent Document 1] Japanese Patent Application Publication No. 2009-296746 [Patent Document 2] Japanese Patent Application Publication No. 2006-288124

In Patent Document 1, it is proposed to install a two-stage notch filter in series in the FB controller system in order to suppress the two resonance characteristics, and use an adaptive notch filter that is connected in parallel to the FB control system. The method of automatic adjustment. Furthermore, for the purpose of distinguishing the notch filter provided in the FB controller system as an adaptive notch filter, hereafter, the notch filter provided in the FB controller system is called a real notch filter.

Specifically, for the motor speed observed by the encoder, two band-pass filters with different band widths are set in parallel (hereinafter, abbreviated as BPF) and applied, and the output of each BPF is adapted to each trap. The wave filter operates, thereby simultaneously estimating the frequency of the vibration component due to the two resonance characteristics of the motor rotation speed, and achieving an automatic adjustment method in a short time by applying the center frequency of each real notch filter.

Moreover, in Patent Document 2, it is proposed to set a plurality of real notch filters in the FB controller system for the purpose of resonance suppression to adapt to the resonance characteristics that change due to long-term deterioration. For this purpose, automatic readjustment of real notch filters is proposed. The method of notch filter.

Specifically, it is to provide a means to estimate the frequency of the vibration from the motor rotation speed observed by the encoder. The frequency of the vibration estimated by this means is judged to be the frequency of the vibration caused by the resonance characteristic that changes due to the deterioration over the years. A plurality of real notch filters are set, and the estimated vibration frequency is compared with the notch frequency of each real notch filter to appropriately identify the real notch filter that should be corrected, thereby automatically suppressing the cause A method based on the resonance phenomenon of the resonance characteristics that change due to long-term deterioration.

In Patent Document 1, if the band width of the two BPFs with respect to the two resonance characteristics is inappropriately set by each BPF for the vibration components caused by the respective resonance characteristics, the expected effect cannot be expected. For example, if one of the two BPFs with similar resonance frequencies of two resonance characteristics extracts the vibration caused by the two resonance characteristics, an adaptive notch filter is used to estimate the two resonance characteristics, resulting in an estimation Error, can not expect the expected effect. Furthermore, there is a problem that it is not easy to appropriately set the band width of the BPF with resonance characteristics.

Furthermore, when the number of resonance characteristics is 3 or more, it is necessary to set the number of BPFs suitable for the resonance characteristics, and it is not easy to appropriately design the band width of each BPF. Moreover, it is necessary to slap beforehand To grasp the number of resonance characteristics of the controlled machine is an issue that requires troublesome adjustments.

Also in Patent Document 2, there is a problem that it is necessary to grasp the number of resonance characteristics of the controlled machine in advance, and there is a problem that it is impossible to cope with the case where two resonance characteristics are degraded at the same time.

Furthermore, when the resonance characteristics have changed with the real notch filter interposed, the frequency of the vibration superimposed on the motor rotation speed may not necessarily resonate with the changed characteristics due to the influence of the intervening real notch filter. The frequency is the same, and the frequency of the vibration superimposed on the motor rotation speed under the set gain of the FB controller does not necessarily coincide with the resonant frequency whose characteristics have changed. This is a problem that the adjustment of the real notch filter may not be smooth.

The purpose of the present invention is to eliminate the need to investigate the number of real notch filters installed in the FB control system and the notch frequency of the real notch filters in advance, and to suppress more than one attributable to the mechanical system with high accuracy and real time. The vibration of the control system’s response due to its resonance characteristics.

A preferred example of the present invention is a motor control device having: a feedback controller that controls a control object including a motor; and a vibration extraction section that extracts the resonance characteristics of one or more control objects caused by the aforementioned control object and superimposes them on One or more vibration components of the response of the control system; the successive frequency estimating unit, which successively estimates the frequency of a certain one of the aforementioned vibration components among the aforementioned vibration components, and uses its output as the vibration frequency estimation Value sequence; and the resonance number estimating unit, which based on the aforementioned vibration frequency estimation value sequence, outputs the number of resonance characteristics that are the cause of vibration superimposed on the response of the control system as a resonance number estimation value sequence, and sets The number of real notch filters corresponding to the value of the estimated resonance number sequence; the output of the feedback controller is given to the current controller through the real notch filter to control the motor.

Another preferable example of the present invention is a real notch filter adjustment method, in which one or more vibration components caused by the resonance characteristics of one or more control objects and superimposed on the response of the control system are extracted. Among the vibration components, the frequency of a certain one of the aforementioned vibration components is successively estimated as the vibration frequency estimation value sequence; based on the aforementioned vibration frequency estimation value sequence, the resonance characteristics that become the cause of vibration superimposed on the response of the control system The number is output as the resonance number estimation value sequence, and the number of real notch filters corresponding to the value of the resonance number estimation value sequence are arranged in series in the back stage of the feedback controller of the control system.

According to the present invention, it is possible to suppress the resonance characteristics of one or more mechanical systems caused by the resonance characteristics of the mechanical system with high accuracy and real time without investigating the number of actual notch filters and the notch frequency of the actual notch filters in advance. Control the vibration of the response of the system.

Hereinafter, the embodiments will be described with reference to the drawings. In addition, in each figure, the same number is attached|subjected to the structural element which has a common function, and the description is abbreviate|omitted. Moreover, in the following, "feedback" is abbreviated as "FB", "notch filter" is abbreviated as "NF", "low-pass filter" is abbreviated as "LPF", and "high-pass filter" is abbreviated as "HPF", abbreviate "Band Pass Filter" as "BPF". [Example 1]

Fig. 1 is a diagram showing a configuration when the automatic adjustment unit 2 of the first embodiment is applied to a general motor FB control system. In a general motor FB control system that does not include the automatic adjustment unit 2, the operation amount of the FB controller 13 is supplied to the motor 14, and the control target machine 15 is controlled via the output y of the motor 14.

The output y is the motor rotation speed [rpm]. The output is measured using a sensor (for example, an encoder). The addition and subtraction calculator 16 calculates the deviation from the rotation speed command r, and the FB controller 13 processes the deviation as a speed deviation. Furthermore, a device (an inverter, etc.) for driving the motor 14 or a controller for controlling the current of the motor 14 is provided in the front stage of the motor 14. These are briefly shown in FIG. 1.

In the FB control system, a notch filter is generally used as a means to suppress vibration or oscillation caused by the resonance characteristic of the control target machine 15. Specifically, for the resonance frequency of the resonance characteristic, in order to make the notch frequency of the notch filter consistent, it is better to set the notch filter in the back stage of the FB controller. Through this, the zero point of the notch filter offsets the resonance pole of the resonance characteristic, and the FB controller 13 can control the control target machine 15 without resonance characteristic excitation (hereinafter, the trap is set in the FB control circuit for the purpose of resonance suppression). The wave filter is called a real notch filter).

The Bode diagram shows a resonance characteristic that is expressed in the transfer characteristic from the motor torque to the motor rotation speed, and the state of canceling it with a real notch filter, as shown in Fig. 16. It is understood that the notch (valley) of the notch filter is used to offset the peak of the resonance characteristic.

It is assumed that there are a plurality of resonance characteristics of the control target machine 15, but the actual notch filter system is only the number of resonance characteristics of the control target machine 15 that is an obstacle to the realization of the desired response characteristic of the FB control system, and it can be set in After the FB controller. By installing a real notch filter, it is possible to reduce the decrease in stability margin due to resonance characteristics in the FB control system, increase the FB control gain, and achieve high response of the FB control system (to achieve the desired response characteristics) .

The automatic adjustment unit 2 is necessary to install the real notch filter 1 to the real notch filter n in the latter stage of the FB controller, thereby suppressing the influence of the resonance of the maximum n of the controlled machine 15 in the FB control system. The automatic adjustment unit 2 is a configuration of a notch filter adjustment device that adjusts the necessary number of real notch filters and the notch frequency of each real notch filter in real time, at high speed, and automatically.

The automatic adjustment unit 2 uses the successive frequency estimation unit 3, the resonance number estimation unit 4, the resonance number judgment unit 5, the vibration extraction unit 6, the vibration detection unit 7, the switch 8, the changeover switch 9, and n real notch filters Constituted. Furthermore, the automatic adjustment unit 2 is based on the premise that it is executed by a digital calculator such as a microcomputer.

The vibration extraction unit 6 takes the output y from the motor as input, extracts the vibration component from y, and outputs the vibration component yd(t). Moreover, yd(t) is matched with the prescribed calculation period Ts of the digital calculator, and outputs yd(0), yd(Ts), yd(2Ts), ....

The automatic adjustment unit 2 aims to match the notch frequency set in the actual notch filter with the resonance frequency. Therefore, it is desirable to extract as much as possible only the vibration component caused by the resonance from the output y. As an example, HPF or BPF is used. The use of LPF is considered from the viewpoint of removing the noise of the sensor that detects the output y, and the use of HPF is considered from the viewpoint of removing the control response that is the constant component from the output y, and extracting only the vibration component.

The filter that satisfies the two viewpoints is LPF+HPF=BPF. These filters are preferably designed to cut off frequencies in accordance with the frequency band to be extracted from the vibration that is the cause of resonance. For example, when the set frequency range of the real notch filter by the automatic adjustment unit 2 is determined to be 100 [Hz] or more, the HPF cut-off frequency is set to 100 [Hz] or the like.

The vibration detection unit 7 takes the output of the vibration extraction unit 6, which is yd(t), as an input. When a significant continuous vibration can be confirmed from yd(t), it acts as an output and takes the generation duration band as 1, except The time zone other than this is used as the task of the vibration detection flag signal of 0.

The initial state of the automatic adjustment unit 2 is a state where even one real notch filter has not been installed in the back stage of the FB controller. In the initial state, when the vibration detection unit 7 does not detect vibration, the vibration detection unit 7 The switch 17 is controlled so that the real notch filter becomes a signal in a state where even one of the real notch filters has not been installed in the latter stage of the FB controller, and is output to the switch 17. In a situation where one or more real notch filters are installed in the back stage of the FB controller, the vibration detection unit 7 switches the switch 17 to make the real notch filters effective. Furthermore, the switching of the switch 17 may be performed by the resonance number estimation unit 4.

The switch 8 takes the output of the vibration detection unit 7, which is the vibration detection flag signal and yd(t) as input, and operates to output yd(t) when the vibration detection flag signal is 1, and output when the vibration detection flag signal is 0. 0.

The successive frequency estimating unit 3 takes the output of the switch 8 as an input, and outputs the vibration frequency estimation value sequence a(k) [Hz].

The successive frequency estimation unit 3 estimates the frequency of the vibration of yd(t) immediately (period Ts) only when the output of the switch 8 is non-zero, and outputs a(k) when the estimation is completed, k=0, 1, .... That is, it should be noted that the vibration frequency estimation value sequence a(k) is not output (updated) every predetermined calculation period Ts, but is output (updated) only when the frequency estimation has been completed.

One thing to note is that the successive frequency estimating unit 3 estimates the frequency of the vibration of yd(t) only when it is judged by the vibration detecting unit 7 that a significant continuous vibration occurs, and the frequency of the vibration of yd(t) is Limited to the frequency band of the vibration that is the cause of resonance that is to be extracted by the vibration extraction unit 6, the vibration frequency estimation value sequence a(k)[Hz] is a non-continuous vibration waveform, not as a vibration or not. The estimated value of the vibration in the significant case.

That is, the vibration extraction unit 6 and the vibration detection unit 7 are designed to prevent the non-continuous vibration of the response of the FB control system caused by impact interference as the estimation target of the successive frequency estimating unit 3, and are responsible for the vibration to be estimated. Give the task of restriction.

From the restriction on yd(t) caused by the vibration extraction unit 6 and the vibration detection unit 7 described above, yd(t) assumes that there are multiple resonance characteristics of the control target machine 15, and the desired result is obtained in the FB control system. The response characteristic superimposes the vibration of the vibration component of several points which is the resonance characteristic of the obstacle.

Assuming that yd(t) is constructed by superimposing n types of vibration components, the successive frequency estimating unit 3 focuses on one vibration component j among n types of vibration components, estimates the frequency of vibration component j, and outputs As a(k).

One of the selection guidelines of j is the vibration with the largest amplitude (power) among the n types of vibration components. Hereinafter, in the present embodiment, the successive frequency estimating unit 3 estimates the frequency of the vibration with the largest amplitude (power) from among the n types of vibration components of yd(t), and outputs it as a(k).

The resonance number estimating unit 4 assumes that there are a plurality of resonance characteristics of the controlled machine 15 based on the vibration frequency estimation value sequence a(k), and obtains the desired response characteristics in the FB control system, and estimates the number of resonance characteristics that become obstacles. , Output the resonance number estimation value sequence N(k), and set the number of real notch filters 1 to n corresponding to the value of N(k) in the latter stage of the FB control system.

The resonance number determination unit 5 takes the vibration frequency estimation value sequence a(k) and the resonance number estimation value sequence N(k) as input, and outputs the number of the real notch filter to which a(k) should be set.

The switch 9 is switched so that it can be set to the actual notch filter to which a(k) should be set in accordance with the number of the actual notch filter obtained from the resonance number determining section 5. Through this, the notch frequency of the real notch filter selected by the switch 9 is updated to a(k).

The above-described processing of the resonance number estimation unit 4, the resonance number determination unit 5, and the changeover switch 9 is repeated every time the vibration frequency estimation value sequence a(k) is updated. In the case of updating the vibration frequency estimation value sequence a(k) as described above, it is limited to the time when a significant continuous vibration caused by resonance occurs. For this reason, when even one real notch filter has a significant continuous vibration that is not caused by resonance in the FB control system, or if there is a(k) at the notch frequency (perform the kth time) In the case where the actual notch filter has not cancelled out the resonance characteristics at the time of the update, the vibration frequency estimation value sequence a(k) is repeatedly and continuously updated. Then, successively set a(k) to the real notch filter. That is, such repeated processing is continued until the real notch filter sufficiently cancels out the resonance characteristics.

The processing flow 20 of such repeated processing is shown in FIG. 2.

The vibration detection unit 7 calculates a vibration detection flag indicating the duration of the period in which the continuous vibration is confirmed, and only when the vibration detection flag is 1, the successive frequency estimating unit 3 estimates the frequency of yd(t).

When the successive frequency estimating unit 3 cannot yet complete the estimation of the vibration frequency estimation value sequence a(k), it continues to estimate the frequency of the continuous yd(t).

When it is judged that the successive frequency estimating unit 3 has completed the estimation of the vibration frequency estimation value sequence a(k), the vibration frequency estimation value sequence a(k) is transmitted through the resonance number estimation unit 4, the resonance number determination unit 5, and the switch 9 is applied to the real notch filter, and the vibration detection unit 7 calculates a vibration detection flag indicating the duration of the period of time when the real notch filter is applied.

If the vibration detection flag is not 1, the processing ends.

The necessity of such repeated processing will be explained using FIG. 3 which is a conceptual diagram showing a convergent plane at the time of resonance. In the FB control system, the frequency ωv of the vibration component yd(t) of the response due to the resonance characteristic does not necessarily coincide with the resonance frequency ωm of the resonance characteristic. In particular, when the FB control gain is increased, the resonance frequency is high, or the delay time in the FB control loop is long, the ωm-ωv system of the two tends to become more significant.

Therefore, the successive frequency estimating unit 3 accurately estimates the frequency of yd(t) and outputs a(k)(=ωv). Even if ωv is applied to the real notch filter as the notch frequency (f2 in Fig. 15), it is not necessarily limited A real notch filter with a(k)(=ωv) can offset the resonance characteristics.

Furthermore, when the real notch filter of ωm≠a(k)(=ωv) is set in the FB control system, the frequency of the vibration component yd(t) of the response due to the resonance characteristics observed is not always ωv . Assuming that the successive frequency estimating unit 3 has been executed while changing to ωv1, the successive frequency estimating unit 3 obtains a(k) as the estimated value of ωv1, but its ωm≠a(k)(=ωv1 ) Is also assumed.

That is, in the case where the obtained a(k) (=ωv1) is set to the real notch filter, as expected, it is not always possible to cancel the resonance characteristics. Therefore, there are many motivations to perform repeated processing as shown in the processing flow 20. In order to allow the real notch filter to cancel the resonance characteristics through repeated processing, a(k) converges to ωm by repeated processing, and a(k)=ωm must be guaranteed.

Now, consider the case where one real notch filter (Figure 15) cancels one resonance characteristic (Figure 16). The transfer characteristic RAR(s) of the resonance characteristic and the transfer characteristic Nch(s) of the notch filter are expressed as follows.

However, ωa, ωm, ζa, and ζm are anti-resonance frequency [rad/s], resonance frequency [rad/s], anti-resonance attenuation coefficient, and resonance attenuation coefficient, respectively. Furthermore, ωn, D, and W are notch frequency [rad/s], notch depth, and notch width, respectively.

The successive frequency estimating unit 3 can accurately estimate the frequency ωv of the vibration component yd(t), a(k)=ωv. With the repeated processing of the processing flow 20, a(k) can converge to the resonance frequency ωm, that is, the equations (3) and (4) are established, and it is expected that d(k) has the same properties as the equation (5) The convergence point list.

Now, let the real notch filter with a(k) (=ωv) set a frequency of the vibration component yd(t) in the latter stage of the FB controller as ωva. At this time, in terms of satisfying the expression (5), it is preferable that the relationship between a(k) and ωva is, for example, CP1 to CP3 in FIG. 3.

That is, the relation of ωva relative to a(k) is to pass through the intersection point coordinates (ωva, a(k))=(ωm, ωm) and do not enter the plane of the oblique line (hereinafter referred to as convergence flat). Also through repeated processing on any convergent plane of CP1 to CP3, its ωm=ωva=a(k).

It is possible to converge to ωm with a small number of repetitions to the extent that the convergence plane is flat. In the case of a convergent plane like CP4, it may not be guaranteed to converge to ωm through repeated processing. However, it may be possible to converge below the initial value of the plane shape or a(k).

The convergence plane is a complex function that changes according to the gain of the FB controller, the delay in the FB control loop, the resonance frequency, the resonance attenuation coefficient, the width of the notch, and the depth of the notch. The convergence of a(k) caused by repeated processing The guarantee of sexual analysis is difficult. Therefore, the method of analyzing convergence caused by repeated processing is limited to grasping a rough phenomenon, and confirming whether the convergence plane satisfies the equations (4) to (5) is supported by the method of determining the value obtained.

Define I, E, and Et as follows.

In equation (6), I and E are the ideal response term and the offset error term when the notch filter is used to cancel the resonance characteristics. From the definition of I, the denominator of E is the resonance characteristic with the resonant pole, and the numerator becomes the real notch filter numerator with the zero point of the real notch filter. When E=1, the zero point of the real notch filter is completely offset. Pin the resonant pole.

On the other hand, it should be noted that I does not include the resonant pole, but does include the notch frequency ωn. As the notch frequency changes (adjustment), the characteristics change. In order to effectively predict the offset error term E, the offset remaining error term Et of equation (8) is defined.

If Et=0, from equation (6), it becomes only the ideal response term I, which can completely cancel the resonance characteristics. Moreover, if the notch depth D=1 is determined, the equation (2) becomes Nch(s)=1. In the case of D=1, the expressions (6)～(8) do not include the actual notch. filter. Therefore, before the iterative process, D=1 is also determined before the real notch filter is installed in the back stage of the FB controller, so that equations (6) to (8) can be analyzed in a unified manner.

In terms of achieving Et=0, the following formula (8) is better.

From the definition of Et, assuming DW=ζm, If |ωn-ωm| decreases monotonously, Et also achieves monotonic Et→0.

Now, take Z(s) as an element other than Nch(s) and RAR(s) in the FB control loop, that is, as the transfer function FB(s) of the FB controller, and the inertia characteristic J(s) of the controlled machine , And the product of the delay characteristic D(s) in the FB control loop is given as follows.

At this time, the closed-loop transmission characteristic of the FB control system (the transmission characteristic of r→y in FIG. 17) is written as follows.

If Et=0, y1 disappears, and the response of the FB control system becomes only the ideal response y0 (=Z(s)I(ωn)/(1+Z(s)I(ωn))). What should be paid attention to is the second term of equation (11). The denominator DE(s) of Et has the resonance characteristic of the resonant pole. Therefore, the second term of equation (11) contains Et in the form of a product, and the second term of equation (11) has a resonance pole. In this case, y1 should contain the vibration component of the resonance frequency ωm.

However, as described above, the frequency system of the vibration of yd(t), which extracts only the vibration component due to resonance from y=y0+y1, does not necessarily coincide with the resonance frequency ωm. It can be grasped by analysing the second term of equation (11) into the following way.

In formula (12), the denominator DE(s) of Et is offset by the DE(s) of the numerator NC(s), y1 does not vibrate at the resonance frequency ωm, and y1 means the root (pole) of DC(s) Vibrate at the frequency of the root (pole) of the resonance cause. This is the reason why the frequency of yd(t) does not coincide with the resonance frequency.

Next, the reason why the frequency of yd(t) is close to the resonance frequency will be explained. y0 is the ideal response in a closed loop system. However, as in equation (11), the denominator of the transfer characteristic of y0 is the same as y1 in the denominator of DC(s), that is, y1 also means the root (pole) that contains DC(s) The components that vibrate at the root (pole) frequency of the resonance cause are not regarded as ideal responses at first glance.

However, the difference between y0 and y1 is that there is a resonance characteristic DE(s) in the transfer characteristic molecule of y0. In other words, it can be explained that among the roots (poles) of DC(s), the roots (poles) of the resonance cause are roughly offset by the resonance characteristic DE(s), and there is almost no effect at y0. This means that the frequency of the vibration of yd(t) (the frequency of the root (pole) of the resonance cause among the roots (poles) of DC(s)) is close to the resonance frequency (the resonance pole of DE(s)).

This means that D=1 (when there is no notch filter) can be said to be common, that is, even if there is a certain degree of difference, regardless of the case of a notch filter The frequency of yd(t) is also close to the resonance frequency.

In addition, when Et approaches 0, the y0 system is idealized as a vibration that substantially does not include the cause of resonance. Moreover, "the roots (poles) of the resonance cause among the roots (poles) of DC(s) are roughly the same as the resonance poles of the resonance characteristics DE(s)." The transfer characteristics of y1 cancel out the resonance of the Et denominator. The DE(s) contained in the NC(s) of the characteristic DE(s) is roughly offset by the DC(s). As a result, the resonance characteristic DE(s) remains in the transfer characteristic of y1.

This means that if Et approaches or coincides with 0, y0 does not include the vibration caused by resonance, and y1 vibrates at the resonance frequency. That is, the frequency of the resonance-induced vibration superimposed on the response of the FB control system becomes the resonance frequency ωm.

Therefore, if Et=0, the intersection coordinates of the convergence plane become (ωm, ωm). Furthermore, in the case of Et=0, from equation (11), y1=0, that is, the response system of the FB control system is y=y0. Therefore, the amplitude of the vibration caused by the resonance of the response y of the FB control system in the vicinity of the intersection coordinates (ωm, ωm) of the convergence plane can be said to be small.

From the equations (11) and (12), the Et approaches or coincides with 0, whereby the frequency of the vibration caused by the resonance coincides with the resonance frequency, and the amplitude is miniaturized. That is, it is understood that the vibration caused by the resonance can be removed, and its purpose can be achieved under the repeated processing described above. In other words, if the notch width W and the notch depth D are appropriately set for the resonance attenuation characteristic ζm, and the notch frequency ωn approaches the resonance frequency ωm, it can be grasped that the purpose can be achieved. Furthermore, it is also possible to grasp that the coordinates of the intersection point of the convergent plane become (ωm, ωm).

However, the convergence plane constructed as shown in Fig. 3 is not a rigorous expression, that is, it is not a rigorous expression of the convergence of repeated processing.

For this reason, the convergent plane that numerically depicts a resonance is shown in Fig. 4. Figure 4 shows the case where the resonance frequency is 1894 [Hz]. Moreover, the setting of FB control gain or delay is to make the setting of FB control system oscillation without real notch filter. The notch width W and notch depth D of the real notch filter are based on the notch frequency and resonance frequency. In the case of coincidence, it is a value that stabilizes the FB control system and does not oscillate.

From the numerical example in Fig. 4, it is understood that the convergence plane does not enter the oblique line part of Fig. 3, and it is in a state where the notch frequency can be aligned with the resonance frequency through repeated processing of the intersection coordinates (ωm, ωm). Furthermore, for example, in a case where the notch width is extremely narrow, etc., there is a case where the convergent plane enters the oblique line portion of FIG. 3.

Therefore, when the width or depth of the actual notch filter is not suitable, even if the processing is repeated, it does not necessarily mean that the notch frequency can be aligned with the resonance frequency. However, if the set value of the actual notch filter is converged If it is suitable for the conditions, in most cases, it can be confirmed by repeated processing to make the notch frequency coincide with the resonance frequency.

In the numerical example in Fig. 4, when the actual notch filter is between approximately 900 and 2000 [Hz], the FB control system is stabilized. This is because when the FB control system oscillates due to resonance, the real notch filter is set to be lower than the resonance frequency, so that the phase advances in the frequency band above the notch frequency of the real notch filter. The characteristic restores the stability margin around the resonance frequency of the FB control system. This is because the resonance frequency ≧ notch frequency easily contributes to the stability of the FB control system.

This means that even if a(k)≠resonance frequency and resonance frequency ≧a(k) in the process of repeated processing, the effect of resonance suppression can be expected. It means that even if a(k)=resonance frequency is not established, the generalization can be obtained. The resonance suppression effect.

So far, it has been explained that there is one resonance characteristic and one real notch filter is used to cancel the resonance characteristic. However, there are n resonance characteristics and n real notch filters are used to cancel these resonance characteristics. The case can also be explained using the analysis method of the above-mentioned equations (6) to (12) and the numerical method.

In order to simplify the description, the case where n=2 is explained. The transfer characteristic RAR(s) of the resonance characteristic and the transfer characteristic Nch(s) of the notch filter are expressed as follows.

As in the case of n=1, the following is defined.

However, I1, Et1, I2, and Et2 are the ideal response term of the first resonance, the offsetting residual term, the ideal response of the second resonance, and the offsetting residual term, respectively. Moreover, the method of determining Dp(p=1, 2)=1 can also be expressed in the situation where there is no real notch filter Nchp(s, ωmp)(p=1, 2).

From equations (16) to (21), the following equations can be obtained, which can clearly express the offset error of Nch1(s,ωn1)･RAR1(s,ωm1), Nch2(s,ωn2)･RAR2(s,ωnm2) Offset errors and the mutual influence of these offset errors.

In terms of achieving Et1=Et2=0, the following formulas (18) and (21) are better.

From the definition of Etp(p=1, 2), assuming that DpWp=ζmp, if |ωnp-ωmp| decreases monotonously, Etp also achieves monotonic Etp→0.

Now, take Z(s) as an element other than Nch1(s), RAR1(s), Nch2(s), and RAR2(s) in the FB control loop, and give equation (10).

At this time, the closed-loop transmission characteristic of the FB control system (the transmission characteristic of r→y in FIG. 18) is written as follows.

The two frequencies of the vibration caused by the resonance overlapping with the response of the FB control system do not coincide with the resonance frequencies ωm1 and ωm2. Also in the case of 2 resonance, y1, y2, and y12 are given the same shape as equation (11). This is because It is the same reason as the aforementioned case of n=1.

In addition, when the convergent plane is drawn, the intersection coordinates (wm1, wm1) and the intersection coordinates (wm2, wm2) must be passed. This is because the reason is the same as the aforementioned case of n=1.

Hereinafter, for convenience, for Nch1(s)･RAR1(s) and Nch2(s)･RAR2(s), use one of them as x and the other as y, written as Nchx(s)･RARx(s) and Nchy (s)･RARy(s).

When the repeated processing is performed in the configuration shown in FIG. 1, if the following conditions are satisfied, two resonance characteristics can be suppressed through the repeated processing.

C1: In each iteration, the real notch filter Nchx is to update and set the resonance characteristic RARx generated by the vibration estimated by the successive frequency estimating unit 3 in a direction that can be more offset C2: In each repetition, the resonance characteristic RARy of the other side of the resonance characteristic RARx caused by the vibration estimated by the successive frequency estimating unit 3 does not reduce the resonance suppression effect of the set real notch filter Nchy･ Neutralize Regarding condition C2, when one of the real notch filters Nchx is updated, the resonance characteristics of the other side RARy and the physical characteristics of the real notch filter Nchy have not changed. The resonance suppression of RARy･Nchy on one side･cancellation effect will not decrease.

However, in the case of a closed loop, since the update of the real notch filter on one side also has some influence on the other side, it is not necessarily guaranteed that the condition C2 is satisfied.

However, according to equation (25), if one of the real notch filters Nchx is updated and the resonance cancellation and suppression effect is improved (that is, Etx approaches 0), the interaction term, Etxy, also approaches 0. , Because the effect of the offsetting residual part (1+Etx+Ety+Etxy) included in all the denominators of y0 to y12 is reduced, y0 approaches the ideal response, which also reduces the offset error in yx and yxy Influence. Therefore, in order to satisfy the update of C1 of the real notch filter Nchx of one of them, it does not significantly reduce the resonance cancellation and suppression effect of the other y.

In particular, when the update of C1 to satisfy one of the real notch filters Nchx becomes Etx≒0, because yx and yxy are approximately 0, there is no need to use the other real notch filter Nchy. Subsequent updates reduce the resonance offset and suppression effect on the x side.

This means that Etx and Ety are close to 0, and the condition C2 is satisfied, which is conducive to convergence. Moreover, when Etx or Ety approaches 0 and approaches 1 resonance (n=1), the condition C1 is satisfied during repeated processing.

It is difficult to prove the rigorous convergence of n=2 or more, and it is supported by numerical methods in the analysis of the convergence plane.

For RAR1 with a resonance frequency of 1894.7 [Hz] and RAR2 with a resonance frequency of 3132.0 [Hz], the FB control system as shown in Fig. 1 is constructed, and two real notch filters Nch1 and Nch2 are drawn from the notch frequency 1100 [Hz]. ] The convergence plane when sliding to 3900 [Hz].

However, because there are two real notch filters, the domain of definition becomes two-dimensional, and the convergent plane becomes a three-dimensional plane. Furthermore, because there are two types of resonance characteristics, there are respective convergent planes for each resonance characteristic. In order to easily determine the evaluation in the 3D plane, slide the real notch filter Nchx of the other side while fixing the real notch filter Nchy of one side to draw the 2D convergence plane, and overlap so that Nchy is fixed at Convergence planes in various situations are plotted and evaluated by this.

Moreover, it is also evaluated that there is no fixed Nchy (that is, Nchy=1). Fig. 5 shows the convergent plane of the first resonance 1894.7 [Hz] using this method, and Fig. 6 shows the convergent plane of the second resonance 3132.0 [Hz].

From Figures 5 and 6, also when the fixed side notch filter Nchy is between an arbitrary frequency, or when it is not between, in the first resonance and the second resonance, it is confirmed that the convergence plane satisfies Formulas (3)～(5). Therefore, 2 resonances can be suppressed by repeated processing.

When the position of the fixed-side notch filter Nchy is near the x-th resonance (for example, in the case where the fixed-side notch filter Nchy is interposed between the second resonance), the convergence plane tends to be flat.

This means that when the resonance characteristic x is more canceled by the notch filter Nchy, it indicates that the intervention frequency of the other notch filter Nchx maintains a high state even if the canceling ability is any one. That is to say, even if it is 2 resonance, the resonance characteristic of the other is more canceled by the notch filter, and furthermore, when it is completely cancelled, the resonance suppression problem can be transferred to the case of 1 resonance. Its system does not contradict the explanation of formula (25).

Similar to the case of n=1, when the width or depth of the actual notch filter is not appropriate, repeated processing may not necessarily make the notch frequency coincide with the resonance frequency, but it will converge to the setting of the actual notch filter. If the value is suitable and other conditions, in most cases, it can be confirmed that the convergence plane satisfies the equations (3) to (5) through numerical methods.

Therefore, even in the case of two resonances, two resonances can be suppressed through repeated processing. In the description so far, the sequence a(k) output by the successive frequency estimating unit 3 is based on the premise that the frequency of the vibration yd(t) (the vibration component with the largest amplitude (power)) can be accurately estimated.

The successive frequency estimating unit 3 for realizing this is shown in FIG. 7. The successive frequency estimating unit 3 is composed of a successive frequency estimator 71, a convergence determiner 72, and an AND process 73. The successive frequency estimator 71 outputs the estimated value of the frequency of the vibration yd(t) at the time t of the vibration yd(t) as the successive frequency estimated value sequence a(t).

The convergence determiner 72 takes a(t) as input, and outputs the convergence determination pulse Pls at the time sequence k (k=0, 1, 2,...) when it is determined that the successive frequency estimation value sequence a(t) has converged to a constant value. (k)k(k=0,1,2,...).

The AND process 73 takes a(t) and Pls(k) as inputs, and outputs the output of the successive frequency estimating unit 3 according to Pls(k), which is the estimated value sequence a(k) (k=0,1,2,...) . The successive frequency estimator 71 is, for example, an adaptive notch filter, an adaptive single-frequency signal booster, or a nonlinear estimator (sine wave adaptor), which can estimate the frequency in real time. The block configuration of the processing in the case where the successive frequency estimator 71 adopts a discrete IIR (Lattice) type adaptive notch filter (1 stage) with a simple configuration is shown in FIG. 8. Next, the adaptive algorithm of the successive frequency estimator 71 is shown below.

＜Discrete IIR notch filter 81＞

＜Adaptation adjuster 82＞

＜Unit Converter 83＞

Furthermore, x, e, and aL are variables that mean internal state quantity, estimated error, and notch frequency, respectively. Furthermore, μ, λ, rL, and σx2 are the update step adjustment coefficient, the forgetting coefficient, the notch width coefficient, and the deviation of x, all of which are positive values. Furthermore, the unit converter 83 converts the unit of aL(t) into [Hz], and outputs it as a(t).

The successive frequency estimation value sequence a(t) of the vibration yd(t) caused by equations (26)～(30), if yd(t) is a vibration waveform with multiple frequency components superimposed, it is The amplitude (power) of a plurality of vibration components is the largest and the tendency to the frequency of the continuous vibration component is estimated with priority (Moreover, when the amplitude (power) ratio of each vibration component approaches 1, the tendency is easy to estimate The initial value a(0) is dependent and close to the vibration component of the frequency of a(0)).

This means that when a(t) is applied as a(k) to the real notch filter and resonance is suppressed, when the adaptive notch filter is applied to the successive frequency estimator 71, the amplitude (power) tends to be the largest Resonance is preferentially suppressed.

Various implementation methods are considered for the convergence determiner 72. However, an example of a simple configuration is shown below.

<Convergence Determinator 72> The following equation defines the difference process.

The output Pls(k) of the convergence determiner is calculated as follows. i) The difference process ε(t) does not exceed the difference threshold Tε even once within the specified time Te, and the absolute value of the difference (slope) between the initial value and the final value of a(t) within the specified time Te When the value is within the slope threshold Tεd, it is judged to be converged, the timing is set to k, and the convergence judgment pulse Pls(k) is set to 1. ii) The difference process ε(t) exceeds the difference threshold Tε at least once within the specified time Te, or after the convergence determination pulse is generated, until the specified time Ted elapses, the convergence determination pulse is zero.

With a simple slope calculation method and a method of setting the slope threshold, when a(t) is constantly increasing slightly, or when a(t) continues to decrease, no convergence judgment is made.

With this, a(k) becomes a reliable estimated value when the convergence of the adaptive algorithm is completed, and the frequency of vibration yd(t) can be expected to be a correct estimated value.

In order to obtain such a(k), it is possible to estimate the resonance frequency based on the multiple resonance characteristics of the aforementioned repeated processing.

On the premise of such a(k), the operation of the resonance number estimation unit 4 when the number of resonance characteristics is at most 2 (n=1, 2) is shown in FIG. 9. The resonance number estimation unit 4 successively estimates the resonance number from a(k) based on a(k), and outputs a resonance number estimation value sequence N(k).

In Fig. 9, in the case of n=1 (the first resonance is F1[Hz]), in the case where a convergent plane as shown in Fig. 4 is formed, a(k) when repeated processing is performed is indicated by a solid line ; In the case of n=2 (the first resonance is F1[Hz], the second resonance is F2[Hz]), in the case of forming a convergent plane as shown in Figures 5 and 6, the dashed lines indicate repeated processing On the occasion of a(k).

However, in the case of n=2, here, the real notch filter 1 is used to suppress the first resonance, and the real notch filter 2 is used to suppress the second resonance. The a(k) system can be correct The estimated frequency value of any resonance characteristic is tentatively determined to be applicable to a suitable real notch filter.

In the case of n=1, the estimated value a(0) at the initial k=0 is the estimated value for the first resonance. However, through the FB control F1≠a(0), it is the result of one iteration of the process. a(1) According to the convergence plane of Fig. 4, one can expect |F1-a(0)|>|F1-a(1)|. According to Figure 4, it should be |F1-a(k)|→F1=a(k)(k→∞), and the next spindle is |a(k-1)-a(k -2)|>|a(k)-a(k-1)|.

On the other hand, in the case of n=2, assuming k=0 and 1, as shown in Fig. 9, when a certain degree of suppression effect is exerted to suppress the first resonance F1, the vibration (power) of the second resonance is compared with that of the first resonance. More significantly, assuming that k=2, a(k) can be obtained for the second resonance.

In this case, not |F1-a(0)|>|F1-a(1)|, but |F1-a(0)|<|F1-a(1)|, not |a(k- 1)-a(k-2)|>|a(k)-a(k-1)|, which is |a(k-1)-a(k-2)|<|a(k)-a( k-1)|The tendency.

Therefore, focusing on the behavior of a(k) in the case of n=1 and n=2, the following simple algorithm can be used to estimate the number of resonances. That is, when the absolute value of the difference between the current value of a(k) obtained from the successive frequency estimating unit 3 and the previous value exceeds a predetermined threshold, the resonance number can be estimated to be 2, and otherwise 1 .

<Resonance number estimation unit 4 (maximum 2 resonance (n=1, 2) case)>

However, Tr and N(k) are the resonance number threshold value [Hz] and the resonance number estimation value sequence, respectively. Also, it should be noted that the successive frequency estimating unit 3 operates when k≧0, and when the vibration detection unit 7 does not detect the vibration, the automatic adjustment unit 2 is in the initial state, as k=-1, and the resonance number is N (-1)=0, that is, not even one real notch filter is installed in the back stage of the controller.

Moreover, when N(k) is 2, it is the aforementioned tentative determination, that is "Real notch filter 1 is used to suppress the first resonance, and real notch filter 2 is used to suppress the second resonance. A(k) can accurately grasp the frequency estimation value of any resonance characteristic. The selection means of the real notch filter applying a(k), that is, the resonance number judging unit 5, can be realized by the following simple algorithm.

<Resonance number judging section 5 (maximum 2 resonance (n=1, 2) case)> Use the following Ln(k) to determine the number of the real notch filter to which a(k) is applied.

However, an1 and an2 are the notch frequency of the real notch filter 1 at time k [Hz] and the notch frequency of the real notch filter 2 at time k [Hz]. .

So far, the resonance number estimating section 4 and the resonance number determining section 5 have been explained when the number of resonances is the maximum of 2 (n=1, 2). However, the resonance number estimating section is also expanded when the number of resonances is 3 or more. 4. It may be the resonance number estimating unit 111 shown in FIG. 10. The algorithm of the resonance number estimation unit 111 is shown below.

Fig. 10 is a diagram showing a modification of the first embodiment of the FB control system applied to the same general motor as in Fig. 1. The description of the same components as in FIG. 1 will be omitted.

ELSE　超過%共振數閾值的情況

IF　a(k)被包含在Rng(1)～Rng(N(k-1))之中 任意一個的Rng(j)的情況

<Resonance Number Estimation Unit 111 (In the case of n resonance)>

ELSE exceeds the% resonance number threshold

When IF a(k) is included in Rng(j) of any one of Rng(1) to Rng(N(k-1))

However, Rng(i) is the i-th resonance bandwidth, which has a specified bandwidth [Wmin(i), Wmax(i)) (that is, Wmin(i)≦Rng(i)<Wmax(i)) By. As an example, when a(k-1) is given, Rng(i) uses WrN which becomes Tr>2×WrN, as Rng(i)=[a(k-1)-WrN, a(k- 1) +WrN).

rN(k) is the resonance bandwidth number, which is the number assigned to each Rng to identify the Rng. The resonance number estimation unit 111 outputs N(k) as the resonance number estimation value sequence, and outputs the resonance bandwidth number rN(k) as the number Ln(k) of the real notch filter to which a(k) is applied.

The above algorithm is to assign the resonance bandwidth and the resonance bandwidth number to a(k-1) when a(k) exceeds the resonance number threshold Tr. Only when a(k) does not belong to any one has been When assigning the resonance bandwidth Rng with the resonance bandwidth number, the algorithm to increase the resonance number by +1. The behavior of the resonance number estimation unit 111 is shown in FIG. 11.

The example in FIG. 11 is the result when there are 1st to 4th resonances, the number of resonances is 4, and the respective resonance frequencies are 550, 1000, 2000, and 4000 [Hz]. Whenever the resonance number threshold is exceeded, the resonance bandwidth Rng is assigned to a(k-1). Only when a(k) does not belong to any existing resonance bandwidth Rng, the resonance number is increased by +1, and finally it can be It is grasped that the estimated value of resonance number N(k) becomes the true value of 4.

The above algorithm is generalized by integrating the resonance number estimation unit 4 and the resonance number judgment unit 5. In other words, the resonance number estimation unit 4 and the resonance number determination unit 5 are those who specialize the above-mentioned algorithm when the resonance number is the maximum of 2, and simplify the processing.

Here, the operation of the resonance number estimation unit 4 will be described. When the vibration is superimposed on the response of the FB control system due to the resonance characteristics of one or more control objects, the resonance number estimating unit 4 sets the initial value of the resonance number estimation value to 1, and furthermore, sets the resonance bandwidth The initial value of the number is set to 1 (step 1).

When the absolute value of the difference between the current value of the vibration frequency estimation value sequence obtained from the successive frequency estimating unit 3 and the previous value exceeds a predetermined threshold (resonance number threshold), the previous value is assigned with a resonance frequency. Wide number of resonance bandwidth (step 2).

The resonance bandwidth is the center value of the previous value. The upper limit is determined by adding a positive specific value to the center value, and the lower limit is determined by subtracting the positive specific value from the center value ( Step 3).

In the case of entering any one of more than one resonance bandwidth that has been set this value, the estimated value of the resonance number is not used, and the resonance bandwidth number is updated to be linked to the resonance bandwidth that includes the current value The resonance bandwidth number (step 4).

In the case where one or more resonance bandwidths of the current value have not been entered, or when one or more resonance bandwidths have not been set, the estimated value of the resonance number is increased by 1, and furthermore, the resonance is increased. The bandwidth number is set to the estimated value of the resonance number that has been increased by 1 (step 5).

If the absolute value of the difference between the current value and the previous value of the vibration frequency estimation value sequence obtained from the successive frequency estimating unit 3 does not exceed a predetermined threshold (resonance number threshold), the resonance number estimation value is not changed, and Do not change the aforementioned resonance bandwidth number (step 6).

The resonance number is estimated by repeating from the aforementioned step 2 to the aforementioned step 6 every time the value is obtained this time, and the output is output as a sequence of the resonance number estimation value.

The resonance bandwidth number is used as the number for applying more than one notch filter in the back stage of the controller, and the current value of the result of successive estimation is applied to the notch filter of the resonance bandwidth number.

The situation of resonance suppression when the automatic adjustment unit 2 shown in FIG. 1 is executed is shown in FIG. 12. Moreover, the resonance number is 2, the first resonance frequency is 1000 [Hz], the second resonance frequency is 2000 [Hz], and the vibration yd(t) is as shown in Figure 12. The state of resonance suppression when two types of vibrations (yd1(t) and yd2(t)) are superimposed on the resonance.

The estimated value of the frequency of the successive frequency estimating section 3 of the vibration yd(t) is always (prioritizes the component with the largest amplitude (power)), but only one is obtained. The method of the judgment unit 5 can accurately suppress the second resonance in a short time under the condition that the second resonance continues to occur.

Of course, also in the case where the two resonances are not simultaneously generated, or when each resonance is continuously generated significantly, it is also possible to immediately suppress each resonance in a short period of time. Moreover, after the vibration (resonance characteristics) to be suppressed is filtered out by the vibration extraction unit 6 and the vibration detection unit 7, the frequency of the vibration yd(t) is estimated by the successive frequency estimator 71, and the convergence determiner 72 is improved as The reliability a(k) of the estimated value of the frequency of the vibration yd(t), as this structure, can achieve high-precision and highly reliable resonance suppression.

According to this embodiment, it is possible to provide a notch filter adjustment device and a motor control device equipped with the same. The notch filter adjustment device is intended to suppress one or more resonance characteristics of the mechanical system without the need for prior investigation The number of real notch filters installed in the FB control system, and the notch frequency of the real notch filter, and also when two or more resonance characteristics occur at the same time, it is automatically and accurately estimated immediately. Adjustment can instantly suppress the resonance characteristics of more than one mechanical system.

Furthermore, in the present embodiment, the automatic adjustment unit 2 uses the motor speed y as an input. However, from the viewpoint of easy extraction of vibration components, the input of the automatic adjustment unit 2 may be used as the adder and subtractor 16 of FIG. 1 The output is the motor speed deviation. Furthermore, from the same viewpoint, the output of the FB controller 13 may also be used.

Moreover, the notch width W and the notch depth D can also be updated in accordance with the update of the notch frequency of the real notch filter caused by a(k).

From the viewpoint of more robust suppression of resonance characteristics, it is better to set the notch width W to be wider, and to set the notch depth to be deeper. However, when a(k) is a low frequency, the notch width is set to be wide, or the notch depth is set to be deep, the characteristics of the notch filter are determined by the notch frequency. The mid-range becomes a tendency for the phase delay of the FB control system to increase, which reduces the stability margin of the FB control system. Depending on the situation, there is a concern that the FB control system will oscillate. For this reason, it is desirable to set the notch width W and the notch depth D according to a(k). Therefore, for example, the notch width W and the notch depth D can be a function of a(k) or MAP as the notch width W(a(k)) and the notch depth D(a(k)). Assign the appropriate value.

Moreover, the notch width W and notch depth D can also be configured to observe and adjust the amplitude of a(k) and vibration yd(t). For example, when a(k) does not change continuously, but the amplitude of vibration yd(t) continues to be large, adjustments such as widening the notch width W or deepening the notch depth D are made.

In addition, the resonance number threshold Tr may be changed according to a(k). That is, Tr(a(k)). It is sufficient to determine an appropriate function, MAP, assuming the mechanical system to be automatically adjusted.

In addition, the resonance bandwidth Rng may be changed as a (k-1). That is, Rng(a(k-1)). It is sufficient to determine an appropriate function, MAP, assuming the mechanical system to be automatically adjusted.

Furthermore, the resonance number estimating unit 111 may set the upper limit to the resonance number to be estimated. Moreover, it is also possible to set a limit to the range of the resonance frequency to be suppressed. For example, when the range of Amin[Hz]～Amax[Hz] is to be suppressed, based on a(t), the convergence determiner 72 that performs the convergence determination can also add "Amin≦a(t)≦ "Amax" is used as the condition for outputting the convergence judgment pulse.

Furthermore, the frequency band extracted by the vibration extraction unit 6 may be limited to Amin [Hz] to Amax [Hz].

In addition, the automatic adjustment unit 2 may also serve as a parameter of the automatic adjustment unit 2 that adjusts the ON/OFF of the operation or the resonance number threshold value, etc., based on the FB gain of the FB controller 13. Among the multiple resonance characteristics of the mechanical system, the FB control system should consider the number of resonance characteristics to be suppressed depends on the FB gain.

Furthermore, for the same reason, the automatic adjustment unit 2 may be configured to adjust the FB gain of the FB controller 13 in accordance with the estimated value of the resonance number or the condition of vibration suppression.

Furthermore, the real notch filter may be in a form other than the expressions (2), (14), and (15). Equations (2), (14), and (15) are continuous systems, and it is necessary to discretize them during installation. However, these filters are discretized by various general z-transforms (ZOH, Tustin transformation, and integrated z-transform). The system is not limited to having the same structure as the equations (26) and (27) of the discrete IIR notch filter 81. Therefore, for example, the discrete IIR notch filter 81 may be used as the real notch filter as it is.

Furthermore, the automatic adjustment unit 2 may additionally be provided with a structure that reproduces the real notch filter. For example, if the number of resonance characteristics to be suppressed is n=2, and the upper limit of the possible use of the real notch filter is 2, in the resonance number estimation unit 4 and the resonance number judgment unit 5, due to any error factors, When a real notch filter Nchx is set to a frequency that is extremely far away from the incorrect frequency of either the first resonance or the second resonance, only the other real notch filter Nchy can definitely suppress the first resonance. The state of resonance with the second. The configuration of detecting such a situation and turning off and resetting the fixed real notch filter Nchx is effective for the robustness of the automatic adjustment unit 2. [Example 2]

The second embodiment is an example of applying the first embodiment to the motor control device, and is an embodiment of the speed control system applied to the cascade B control system of the AC servo motor shown in FIG. 13. The control system shown in FIG. 13 includes: an addition and subtraction calculator 1312, a speed controller 132, a current controller 133, a first coordinate converter 134 for converting from a dq coordinate system to a 3-phase coordinate system, and from a 3-phase coordinate system coordinate Converted to the second coordinate converter 1310 of the dq coordinate system, the PWM output 135 that inputs the 3-phase voltage command and outputs the PWM pulse, the inverter (power converter) 136, the current detector 138, the position and speed calculation unit 1311 An encoder 139, a motor 137, and a machine 1313 driven by the motor that measure the rotation speed of the motor.

Fig. 14 shows the application of the automatic adjustment unit 2 shown in Fig. 1 to the second embodiment of Fig. 13. The automatic adjustment unit 1401 takes the motor speed (motor rotation speed) calculated by the output position/speed calculation unit 1311 of the encoder 139 as an input. The output of the speed controller 132 is given to the current controller 133 through the real notch filter to control the motor 137.

The current controller 133 controls the circuit part of the motor. Under the premise that the control cycle is faster than the speed controller 132, in the speed control system, the current control system is considered to be approximately 1 (the operation amount of the speed controller directly reaches the mechanical machinery of the motor). Part (rotor)). Therefore, the control object of the speed controller 132 is the machine 1313 that is combined with the mechanical part (rotor) of the motor and the motor rotor, which is equivalent to the control object of the FB controller in FIG. 1; the speed controller inputs addition and subtraction The output of the arithmetic unit 1312, and the addition and subtraction arithmetic unit calculates the deviation between the output of the position/speed calculation unit 1311 and the rotation speed command.

The inertia number of the machine 1313 is 1, and the machine 1313 and the motor rotor are regarded as elastically coupled, the control target system can be regarded as a two-inertia system that combines the machine 1313 and the motor rotor with a spring and a damper, and the control target becomes a system that includes 1 The frequency characteristics of the resonance and anti-resonance characteristics of the group.

Moreover, the number of inertia of the machine 1313 is 2, and each inertial system is combined with a spring and a damper, and one of them is regarded as an elastic connection to the motor rotor, and the control target system can be regarded as 3 of the inertia combined with a spring and a damper. An inertial system has a frequency characteristic including two sets of resonance and anti-resonance characteristics.

As shown in Example 1, the automatic adjustment unit 2 does not need to investigate the number of resonances in advance, and even if it is 2 resonances or more, resonance can be automatically suppressed. Therefore, also in this embodiment, the automatic adjustment unit 2 does not need to investigate the resonance number in advance, and can automatically set and adjust the actual notch filter including the appropriate notch frequency in the latter stage of the speed controller 132. .

Therefore, according to this embodiment, the automatic adjustment unit 2 is also applied to the speed control system in the AC servo motor serial FB control system shown in FIG. 13, and there is no need to investigate the actual notch filter installed in the speed control system in advance. The number of notch filters and the notch frequency of the actual notch filter, and even when two or more resonance characteristics are generated at the same time, it can be highly accurate, automatically estimated and adjusted in real time, and can instantly suppress more than one of the mechanical systems Resonance characteristics. Furthermore, it is possible to provide a motor control device provided with a series B control system including an AC servo motor of such an automatic adjustment unit 2.

The above-mentioned embodiments can be applied to, for example, semiconductor inspection devices, main motor control devices of electric vehicles, electric power steering, etc., in addition to motor control devices.

2: Automatic adjustment part 3: Successive frequency estimation section 4: Resonance number estimation section 5: Resonance number judgment section 6: Vibration extraction part 7: Vibration detection department 8: switch 9: Toggle switch 10~12: Real notch filter 1~n 13: FB controller 14: Motor 15: Controlled machinery

Fig. 1 is a diagram showing the first embodiment of the FB control system applied to a general motor.

[Fig. 2] is the processing flow of the repeated processing in the first embodiment.

[Figure 3] is a conceptual diagram showing a convergent plane at resonance.

[Figure 4] is a graph that numerically plots a convergent plane at resonance.

[Fig. 5] A diagram showing the convergence plane of the first resonance.

[Fig. 6] A diagram showing the convergence plane of the second resonance.

[Fig. 7] A diagram showing the successive frequency estimation unit.

[Figure 8] is a block diagram of the successive frequency estimator.

[Fig. 9] A diagram showing the operation of the resonance number estimation unit.

[Fig. 10] is a diagram showing a modification of Fig. 1.

[Fig. 11] A diagram showing the behavior of the resonance number estimation unit.

[Fig. 12] A diagram showing the situation of resonance suppression when the automatic adjustment unit is executed.

[Figure 13] is a diagram showing the speed control system of the AC servo motor.

[Fig. 14] A diagram showing Example 2. [Fig.

[Fig. 15] is a graph showing the cancellation of 1 resonance characteristic with 1 real notch filter.

[Figure 16] is a graph to look at the frequency characteristics using a Bode diagram.

[Fig. 17] A diagram showing the closed-loop transmission characteristics (part 1) of the FB control system.

[Figure 18] A diagram showing the closed-loop transmission characteristics (Part 2) of the FB control system.

2: Automatic adjustment part

3: Successive frequency estimation section

4: Resonance number estimation section

5: Resonance number judgment section

6: Vibration extraction part

7: Vibration detection department

8: switch

9: Toggle switch

10~12: Real notch filter 1~n

13: FB controller

14: Motor

15: Controlled machinery

16: Addition and subtraction operator

17: switch

Claims (14)

A motor control device including: a feedback controller that controls a control object including a motor; and a vibration extraction unit that extracts the resonance characteristics of one or more control objects that are caused by the control object and overlaps the response of the control system. The successive frequency estimation unit, which is based on the aforementioned vibration components, successively estimates the frequency of a certain aforementioned vibration component, and uses its output as a sequence of vibration frequency estimation values; and a resonance number estimation unit, which is based on the aforementioned vibration frequency In the estimated value sequence, the number of resonance characteristics that is the cause of the vibration superimposed on the response of the control system is output as the resonance number estimated value sequence, and the number of real values corresponding to the value of the resonance number estimated value sequence is set Notch filter; the output of the feedback controller is given to the current controller through the real notch filter to control the motor. The motor control device according to claim 1, wherein the successive frequency estimating unit has: a successive frequency estimator that estimates the frequency of the vibration component with the largest amplitude from one or more of the vibration components, and outputs it as the successive frequency estimator A sequence of estimated frequency values; and a convergence determiner, which determines whether the sequence of successive frequency estimates has converged to a constant value based on the sequence of successive frequency estimates; The value of the successive frequency estimation value series is the estimated value series, and the estimated value series is determined as the vibration frequency estimation value series as the output of the successive frequency estimation unit. The motor control device according to claim 1, which has a resonance number determination unit that is based on the resonance number of the current time of the resonance number estimation value sequence estimated by the resonance number estimation unit and the aforementioned successive frequency estimation The frequency of the current time of the vibration frequency estimation value sequence estimated by the part, for one or more real notch filters installed in the subsequent stage of the feedback controller, select the frequency of the current time of the vibration frequency estimation value sequence Corresponding to the real notch filter. The motor control device of claim 1, wherein the resonance number estimation unit is based on the vibration frequency estimation value sequence obtained from the successive frequency estimation unit. The absolute value of the difference between the current value and the previous value exceeds the resonance number In the case of a threshold, the number of resonance characteristics is estimated to be two, and if the resonance number threshold is not exceeded, the number of resonance characteristics is estimated to be one, and the estimated number of resonance characteristics is output as the resonance number Predicted value sequence. Such as the motor control device of claim 1, wherein the aforementioned resonance number estimation unit sets the initial value of the resonance number estimation value to 1, and furthermore, sets the initial value of the resonance bandwidth number to 1, As process A, when the absolute value of the difference between the current value and the previous value of the vibration frequency estimation value sequence obtained from the successive frequency estimating unit exceeds the resonance number threshold, the previous value is assigned to the resonance The resonance bandwidth corresponding to the bandwidth number. The aforementioned resonance bandwidth is based on the previous value as the center value, and the value after adding a positive specific value to the aforementioned center value is used as the upper limit, and the positive value is subtracted from the aforementioned center value. The value after the specific value is used as the lower limit, that is, the frequency range is determined, and the estimated value of the resonance number is not changed when it enters any one of the above-mentioned resonance bandwidths where one or more of the above-mentioned current values have been set , And update the aforementioned resonant bandwidth number to the aforementioned resonant bandwidth number that is linked to the aforementioned resonant bandwidth including the aforementioned value of this time. In the case, or if one or more of the aforementioned resonance bandwidths are not provided, the aforementioned resonance number estimation value is increased by 1, and further, the aforementioned resonance bandwidth number is set to increase the aforementioned resonance number estimation value by 1 Value, as process B, if the absolute value of the difference between the current value and the previous value of the vibration frequency estimation value sequence obtained from the successive frequency estimating unit does not exceed the resonance number threshold, the resonance is not allowed The estimated value of the number changes without causing the aforementioned resonance bandwidth The number is changed, and each time the current value of the vibration frequency estimation value sequence is obtained, the resonance number is estimated by repeating the aforementioned processing A and the aforementioned processing B successively, and output as the aforementioned resonance number estimation value sequence. For example, the motor control device of claim 5, wherein more than one real notch filter is provided after the feedback controller of the control system is assigned a positive number from 1 liter power; the resonance bandwidth number is based on the aforementioned The aforementioned current value of the aforementioned vibration frequency estimation value sequence estimated by the successive frequency estimation unit is used as the number applicable to the aforementioned real notch filter installed at the rear stage of the feedback controller of the aforementioned control system, and the aforementioned vibration frequency estimation value The aforementioned current value of the sequence is applied to the aforementioned real notch filter with the aforementioned resonance bandwidth number. The motor control device according to claim 1, wherein, in the initial state, the vibration detection unit does not determine that there is vibration even once from the start of the operation, the actual trap is not installed after the feedback controller of the control system Wave filter. A real notch filter adjustment device has: a vibration extraction unit that extracts one or more vibration components that are caused by the resonance characteristics of one or more control objects and are superimposed on the response of the control system; and a successive frequency estimating unit, which In the aforementioned vibration components, the frequency of one of the aforementioned vibration components is successively estimated, and its output is used as the vibration frequency estimation Value sequence; and the resonance number estimating unit, which outputs the number of resonance characteristics that are the cause of vibration superimposed on the response of the control system based on the aforementioned vibration frequency estimation value sequence, and outputs as a resonance number estimation value sequence, and The number of real notch filters corresponding to the values of the aforementioned resonance number estimation value sequence are arranged in series in the back stage of the feedback controller of the aforementioned control system. The real notch filter adjustment device of claim 8, wherein the successive frequency estimating unit has: a successive frequency estimator that estimates the frequency of the vibration component with the largest amplitude from one or more of the vibration components, And output as a sequence of successive frequency estimation values; and a convergence determiner, which determines whether the sequence of successive frequency estimation values has converged to a constant value based on the aforementioned sequence of successive frequency estimation values; the aforementioned convergence determiner is every time it determines that it has converged, The value of the successive frequency estimation value sequence at that time is output as an estimation value sequence, and the estimation value sequence is determined as the vibration frequency estimation value sequence as the output of the successive frequency estimation unit. The real notch filter adjustment device according to claim 8, which has a resonance number judgment unit that is based on the resonance number of the current time of the resonance number estimation value sequence estimated by the resonance number estimation unit, and The current time frequency of the vibration frequency estimation value sequence estimated by the successive frequency estimating unit is set in the latter stage of the feedback controller For one or more real notch filters, a real notch filter corresponding to the frequency of the aforementioned current time of the aforementioned vibration frequency estimation value sequence is selected. The real notch filter adjustment device of claim 8, wherein the resonance number estimation unit is the absolute value of the difference between the current value and the previous value of the vibration frequency estimation value sequence obtained from the successive frequency estimation unit If the resonance number threshold is exceeded, the number of resonance characteristics is estimated to be two. If the resonance number threshold is not exceeded, the number of resonance characteristics is estimated to be one, and the estimated number of resonance characteristics is output. As the aforementioned resonance number estimation value sequence. A real notch filter adjustment method, in which one or more vibration components that are caused by the resonance characteristics of one or more control objects and overlap the response of the control system are extracted, and among the vibration components, one of the aforementioned vibration components is successively estimated The frequency of the vibration component is taken as the vibration frequency estimation value sequence; according to the aforementioned vibration frequency estimation value sequence, the number of resonance characteristics that is the cause of the vibration superimposed on the response of the control system is output as the resonance number estimation value sequence, In addition, a number of real notch filters corresponding to the value of the aforementioned resonance number estimation value sequence are arranged in series in the rear stage of the feedback controller of the aforementioned control system. Such as the real notch filter adjustment method of claim 12, Among them, the aforementioned real notch filter adjustment method is in the initial state; if the aforementioned real notch filter adjustment method does not detect the vibration caused by resonance even once from the beginning of the process, it is not set at the back of the feedback controller of the aforementioned control system The aforementioned real notch filter. A motor control method in which the real notch filter adjustment method described in claim 12 is used.

Images