Control assistance device, control system and filter adjustment method

Abstract

While achieving resonance suppression, allocation of filters is determined in a manner that satisfies a constraint on the number of filters. A control assistance device that performs assistance for setting a plurality of first filters provided in a servo control device that controls a motor, the control assistance device including: a resonance detection section that detects a plurality of resonance points in a frequency characteristic of an input-output gain and a phase lag of an input and an output of the servo control device measured from an input signal and an output signal that vary with frequency; a filter setting section that sets a plurality of second filters that exceed the number of the plurality of filters in order to suppress the plurality of resonance points; and a grouping section that groups the plurality of second filters to set the plurality of first filters.

Description

Technical Field

[0001] The present invention relates to a control auxiliary device for assisting in setting multiple filters in a servo control device for controlling an electric motor, a control system including the control auxiliary device and the servo control device, and a filter adjustment method. Background Technology

[0002] In machinery with multiple resonant points, in order to maintain stability and improve feedback characteristics, it is necessary to use multiple filters to suppress multiple resonant points.

[0003] In a machine with multiple resonant points, for example, Patent Document 1 describes a control system equipped with a servo control device that suppresses multiple resonant points through multiple filters and a machine learning device that optimizes the coefficients of the filters.

[0004] Patent Document 1 describes a control system in which, when multiple resonant points exist in a machine, multiple filters are installed in a servo control unit (called a servo control device) and connected in series, corresponding to each resonant point, thereby attenuating all resonants. Furthermore, Patent Document 1 describes a machine learning device that sequentially calculates the optimal values ​​for attenuating the resonant points by using machine learning to evaluate the coefficients of the multiple filters.

[0005] Existing technical documents

[0006] Patent documents

[0007] Patent Document 1: Japanese Patent Application Publication No. 2020-57211 Summary of the Invention

[0008] The problem that the invention aims to solve

[0009] As mentioned above, in a machine with multiple resonant points, it is necessary to use multiple filters to suppress multiple resonant points, but the number of filters that can be set in a servo control device is limited.

[0010] In order to suppress multiple resonant points, it is important to decide how to allocate a limited number of filters, but it is difficult to blindly search and decide the allocation of filters while suppressing resonance through machine learning or optimization algorithms.

[0011] Therefore, it is desirable to have a control auxiliary device, a control system, and a filter adjustment method that can achieve resonance suppression and determine the filter allocation in a way that satisfies the constraint of the number of filters.

[0012] Methods for solving problems

[0013] (1) A first aspect of this disclosure is a control auxiliary device that assists in setting a plurality of first filters, the plurality of first filters being disposed in a servo control device for controlling an electric motor, wherein the control auxiliary device comprises: a resonance detection unit that detects a plurality of resonance points in the frequency characteristics of the input-output gain of the servo control device, which are determined based on input and output signals that change in frequency; a filter setting unit that sets a plurality of second filters in order to suppress the plurality of resonance points, exceeding the number of the plurality of first filters; and a grouping unit that groups the plurality of second filters to set the plurality of first filters.

[0014] (2) A second aspect of this disclosure is a control system comprising: a servo control device that controls a motor; and a control auxiliary device as described in (1) above, which detects multiple resonant points in the frequency characteristics of the input-output gain and the phase lag of the input-output of the servo control device, and sets multiple second filters exceeding the number of multiple first filters set in the servo control device in order to suppress the multiple resonant points, and groups the multiple second filters to set the multiple first filters.

[0015] (3) The third aspect of this disclosure is a filter adjustment method, which is a filter adjustment assistance method for a control auxiliary device, wherein the control auxiliary device assists in setting a plurality of first filters, the plurality of first filters being disposed in a servo control device for controlling a motor, wherein the filter adjustment method performs the following processing: detecting a plurality of resonant points in the frequency characteristics of the input-output gain of the servo control device, which are determined based on the input signal and output signal of the frequency change; setting a plurality of second filters, exceeding the number of the plurality of first filters, in order to suppress the plurality of resonant points; and grouping the plurality of second filters to set the plurality of first filters.

[0016] Invention Effects

[0017] According to the various methods of this disclosure, the allocation of filters can be determined in a way that satisfies the constraint of the number of filters while achieving resonance suppression. Attached Figure Description

[0018] Figure 1 This is a block diagram illustrating the control system according to the first embodiment of the present disclosure.

[0019] Figure 2 This is a block diagram illustrating an example of a filter being constructed by directly connecting multiple filters.

[0020] Figure 3 This is a diagram showing the gain characteristics of the input and output gains, the filter settings for multiple resonant points, and an example of grouping.

[0021] Figure 4 This is a block diagram showing the structure of the control auxiliary unit including the frequency characteristic estimation unit in the first embodiment.

[0022] Figure 5 It is a block diagram representing the canonical model used to calculate the input-output gain.

[0023] Figure 6 It is a characteristic diagram showing the frequency characteristics of the input and output gain of the servo control unit in the standard model and the estimated values ​​of the frequency characteristics of the input and output gain of the servo control unit.

[0024] Figure 7 This is a diagram showing the gain characteristics of the input and output gains, the filter settings for multiple resonant points, and an example of grouping.

[0025] Figure 8 It is a block diagram representing the structure of a part of a control system centered on a grouping unit.

[0026] Figure 9 This is a block diagram illustrating the machine learning unit of one embodiment of the present invention.

[0027] Figure 10 It means Figure 1 The flowchart shown illustrates the operation of the control auxiliary unit.

[0028] Figure 11 This is a block diagram illustrating the control system of the second embodiment of the present disclosure.

[0029] Figure 12 This is a block diagram representing a variation of the control system. Detailed Implementation

[0030] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings.

[0031] (First Implementation)

[0032] Figure 1 This is a block diagram illustrating the control system according to the first embodiment of the present disclosure.

[0033] The control system 10 includes a servo control unit 100, a frequency generation unit 200, a frequency response measurement unit 300, and a control auxiliary unit 400. The servo control unit 100 corresponds to a servo control device for controlling a motor, the frequency response measurement unit 300 corresponds to a frequency response measurement device, and the control auxiliary unit 400 corresponds to a control auxiliary device.

[0034] Furthermore, one or more of the frequency generation unit 200, the frequency response measurement unit 300, and the control auxiliary unit 400 may be provided within the servo control unit 100. The frequency response measurement unit 300 may also be provided within the control auxiliary unit 400.

[0035] The servo control unit 100 includes a subtractor 110, a speed control unit 120, a filter 130, a current control unit 140, and a motor 150. The subtractor 110, speed control unit 120, filter 130, current control unit 140, and motor 150 constitute a servo system with a speed feedback loop.

[0036] The control auxiliary unit 400 detects multiple resonant points of the frequency response of the input-output gain (amplitude ratio) of the servo control unit 100, calculated without filter 130. It sets multiple filters exceeding a limit on the number of filters to suppress resonance at these multiple resonant points, and groups these multiple filters to a number below the limit. Filter 130 is set based on the combination of the grouped filters. Figure 1 The image shows a servo control unit 100 after a filter 130 configured by grouping has been set.

[0037] The electric motor 150 is a linear motor that performs linear motion or a motor with a rotating shaft, etc. The object driven by the electric motor 150 is, for example, a machine tool, a robot, or a mechanism of industrial machinery. The electric motor 150 may also be integrated into a machine tool, robot, or industrial machinery. The control system 10 may also be integrated into a machine tool, robot, or industrial machinery.

[0038] The subtractor 110 calculates the difference between the input speed command and the detected speed from the speed feedback, and outputs this difference as a speed deviation to the speed control unit 120.

[0039] The speed control unit 120 performs PI (Proportional-Integral) control, adding the value obtained by multiplying the speed deviation by the integral gain K1v and integrating it, to the value obtained by multiplying the speed deviation by the proportional gain K2v, and outputting this as a torque command to the filter 130. The speed control unit 120 includes a feedback gain. Furthermore, the speed control unit 120 is not particularly limited to PI control; other control methods, such as PID (Proportional-Integral-Differential) control, can also be used.

[0040] Mathematical expression 1 (hereinafter referred to as numerical expression 1) represents the transfer function G of the speed control unit 120. V (s).

[0041] [Formula 1]

[0042]

[0043] Filter 130 is constructed by connecting multiple filters in series to attenuate specific frequency components. Each filter is, for example, a notch filter, a low-pass filter, or a band-stop filter. In machines such as machine tools having a mechanism driven by an electric motor 150, multiple resonant points sometimes exist, and sometimes these resonants increase in the servo control unit 100. By connecting filters such as notch filters in series, the resonants at multiple resonant points can be reduced. The output of filter 130 is output as a torque command to the current control unit 140.

[0044] Figure 2 This is a block diagram illustrating an example of directly connecting multiple filters to form filter 130. Figure 2 In the case where there are k resonant points (k is a natural number greater than 2), filter 130 is constructed by connecting m filters 130-1 to 130-m (m is a natural number greater than 2, m < k) in series. The m filters 130-1 to 130-m correspond to different frequency bands. The following explanation assumes that filter 130 is composed of m filters 130-1 to 130-m.

[0045] Regarding the number m of filters constituting filter 130, the number of filters that can be set by the servo control unit 100 is limited, and the number of filters is constrained, so it is difficult to set the same number as k resonant points. If the constraint on the number of filters is set to the number of filters Nmax, then the control auxiliary unit 400 determines the allocation of m filters for the k resonant points in such a way that the number of filters m is less than or equal to the number of filters Nmax.

[0046] Mathematical expression 2 (hereinafter referred to as expression 2) represents the transfer function G of one of the notch filters in filter 130, for example, filter 130-1. F (s). Filters 130-2 to 130-m can also be constructed from notch filters with the same transfer function.

[0047] Here, the coefficient δ in equation 2 is the attenuation coefficient, and the coefficient ω c ω is the center angular frequency, and the coefficient τ is the relative bandwidth. If the center frequency is set to fc and the bandwidth to fw, then the coefficient ω... c By ω c =2πfc represents the coefficient τ, which is represented by τ = fw / fc.

[0048] [Formula 2]

[0049]

[0050] The current control unit 140 generates a voltage command for driving the motor 150 based on the torque command, and outputs the voltage command to the motor 150.

[0051] When the motor 150 is a linear motor, the position of the movable part is detected by a linear scale (not shown) provided on the motor 150. The speed detection value is obtained by differentiating the position detection value, and the obtained speed detection value is input to the subtractor 110 as speed feedback.

[0052] In the case where the motor 150 is a motor with a rotating shaft, the rotation angle position is detected by a rotary encoder (not shown) installed on the motor 150, and the speed detection value is input to the subtractor 110 as speed feedback.

[0053] The servo control unit 100 is configured as described above.

[0054] In addition to the servo control unit 100, the control system 10 also includes a frequency generation unit 200, a frequency response measurement unit 300, and a control auxiliary unit 400. The frequency response measurement unit 300 may also be included in the control auxiliary unit 400.

[0055] The frequency generation unit 200 outputs a sine wave signal as a speed command to the subtractor 110 of the servo control unit 100 and the frequency characteristic measurement unit 300 while changing the frequency.

[0056] The frequency response measurement unit 300 uses a speed command (sine wave) generated by the frequency generation unit 200 as an input signal and a detection speed (sine wave) output from a rotary encoder (not shown) as an output signal, and measures the amplitude ratio (input-output gain) and phase lag of the input signal to the output signal at each frequency specified by the speed command. Alternatively, the frequency response measurement unit 300 uses a speed command (sine wave) generated by the frequency generation unit 200 as an input signal and a derivative (sine wave) of the detection position output from a linear scale, and measures the amplitude ratio (input-output gain) and phase lag of the input signal to the output signal at each frequency specified by the speed command.

[0057] When the servo control unit 100 operates without the filter 130 installed, the frequency response measurement unit 300 measures at least the frequency response characteristics of input / output gain and phase hysteresis, and outputs the frequency response characteristics of input / output gain and phase hysteresis to the resonance detection unit 401 and frequency response estimation unit 403 of the control auxiliary unit 400 (described later). When the servo control unit 100 operates with some or all of the m filters 130-1 to 130-m equipped with the filter 130, the frequency response measurement unit 300 measures the frequency response characteristics of input / output gain and phase hysteresis, and outputs the frequency response characteristics of input / output gain and phase hysteresis to the grouping unit 404 of the control auxiliary unit 400 (described later).

[0058] like Figure 3 As shown, the servo control unit 100 operates without the filter 130 set, and the control auxiliary unit 400 detects k resonant points of the input-output gain output from the frequency response measurement unit 300. Then, the control auxiliary unit 400, for example, performs n operations exceeding the constraint of the number of filters, i.e., the filter number Nmax. ini The coefficients of each filter (the coefficients ω in mathematical formula 2) c Initial adjustments of (τ, δ), and n ini The filters are grouped, and the number of filters n is set compared to the number of filters initially adjusted. ini The number of filters m(n) is less than or equal to the number of filters Nmax. ini A filter 130 (>m, Nmax≥m) is used to suppress the resonance at the detected k resonant points. The m filters are multiple first filters, and n... ini Each filter is a second filter.

[0059] Number n ini The number can be less than k, more than k, or the same. That is, among k resonant points, multiple filters can be set for one resonant point, one filter can be set for multiple resonant points, or k filters (n) can be set for k resonant points. ini =k). Furthermore, when the number of resonant points k is less than or equal to the number of filters Nmax, sometimes n is also required. ini The initial adjustment of each coefficient of the filter. This is because, when multiple filters are set for one resonant point, n ini >k, there exists a filter number n ini The case where the number of filters exceeds Nmax.

[0060] The number n filters configured to suppress resonance at k resonant points ini The number of filters that exceeds the limit of the number of filters, i.e., the maximum number of filters Nmax, but in the case of the number of filters n iniWhen the number of filters is less than or equal to Nmax, n can also be performed. ini The number of filters is set to m(n) when the number of filters is less than or equal to Nmax. ini The filter has 130 filters (>m). This is because, even if the objective function, such as the cutoff frequency, deteriorates, there is still a situation where fewer filters are better.

[0061] Figure 3 This is a diagram showing the gain characteristics of the input and output gains, the filter settings for multiple resonant points, and an example of grouping. Figure 3 The following states are shown: Eight resonant points P1 to P8 are generated, and the coefficients of the eight filters are initially adjusted in a way that suppresses the resonant points P1 to P8. One filter is set for each of the resonant points P2 and P3, P5 and P6, and P7 and P8 by grouping.

[0062] The structure and operation of the control auxiliary unit 400 will be described in detail below.

[0063] <Control Auxiliary Unit 400>

[0064] like Figure 1 As shown, the control auxiliary unit 400 includes a resonance detection unit 401, a filter setting unit 402, a frequency characteristic estimation unit 403, and a grouping unit 404.

[0065] When the servo control unit 100 operates without the filter 130 set, the resonance detection unit 401 obtains the frequency characteristics of the input-output gain (amplitude ratio) of the servo control unit 100 from the frequency characteristic measurement unit 300, and detects k resonance points of the frequency characteristics (gain characteristics) of the input-output gain.

[0066] In addition to detecting the resonant point, the resonance detection unit 401 can also detect the anti-resonant point. By detecting the anti-resonant point, n is set by the filter setting unit 402. ini n ini Given the range of attenuation center frequencies of each filter in the >m) filter, the range of attenuation center frequencies can be set between the frequencies of the anti-resonance point.

[0067] The filter setting unit 402 sets n corresponding to the k resonant points. ini Each filter, in conjunction with the frequency response estimation unit 403, initially adjusts n according to the filter. ini The coefficients of each filter.

[0068] Grouping section 404 adjusts the initial n ini The filters are grouped into m filters (k > m).

[0069] The following describes the initial adjustment of the filter performed by the filter setting unit 402 and the grouping performed by the grouping unit 404.

[0070] As mentioned above, the number of filters n to be initially adjusted ini It can exceed the maximum number of filters, Nmax, or be less than Nmax. Furthermore, the number of filters n... ini The number of filters m can exceed the number of resonant points k, or it can be less than the number of resonant points k. The number of filters m is less than or equal to the number of filters Nmax, and less than the number of filters n. ini few.

[0071] In the following description, the number of filters to be initially adjusted is n. ini The number of filters Nmax exceeds the number of resonant points k (n ini >Nmax, n ini =k).

[0072] Since the number of filters n to be initially adjusted ini Since the filter quantity Nmax is exceeded, the frequency response estimation unit 403 is used to estimate the frequency response of the input and output gain in the following description.

[0073] However, when the number of filters n to be initially adjusted ini If the number of filters Nmax is not exceeded, even if the frequency characteristic of the input and output gain is estimated by the frequency characteristic estimation unit 403, the servo control unit 100 can actually operate, and the frequency characteristic of the input and output gain can be measured by the frequency characteristic measurement unit 300.

[0074] (Initial adjustment of the filter)

[0075] The following explains how to apply the filter to k filters (k = n) for each filter. ini The coefficient ω c The operation of the filter setting unit 402 that performs initial adjustments to τ and δ, and the structure and operation of the frequency characteristic estimation unit 403.

[0076] The filter setting unit 402 sequentially selects one resonant point from k resonant points, and sets the coefficients ω of the filter to attenuate the resonance of the selected resonant point. c , τ, δ, and are sequentially output to the frequency response estimation unit 403.

[0077] The frequency response estimation unit 403 estimates the coefficients ω of a filter obtained from the filter setting unit 402. cThe frequency characteristics of the input-output gain and phase hysteresis obtained from the frequency characteristic measurement unit 300 when the servo control unit 100 operates without the filter 130 are calculated. The estimated value of the frequency characteristics of the input-output gain of the servo control unit 100 when the filter 130 is set is then output to the filter setting unit 402.

[0078] Figure 4 This is a block diagram showing the structure of the control auxiliary unit, which includes a frequency response estimation unit. Figure 4 In order to illustrate the operation of the frequency response estimation unit 403, the frequency response measurement unit 300 is also shown.

[0079] like Figure 4 As shown, the frequency response estimation unit 403 includes a filter information acquisition unit 4031, a pre-adjustment state storage unit 4032, a frequency response calculation unit 4033, and a state estimation unit 4034.

[0080] The filter information acquisition unit 4031 acquires the coefficients ω of one filter from the filter setting unit 402. c τ, δ are output to the frequency response calculation unit 4033.

[0081] As described above, the frequency characteristics C1 of input / output gain and phase hysteresis obtained from the frequency characteristic measurement unit 300 are stored in the pre-adjustment state storage unit 4032 when the servo control unit 100 operates in a state where the filter 130 is not set.

[0082] The frequency response calculation unit 4033 obtains the coefficients ω of a filter from the filter information acquisition unit 4031. c , τ, δ, and read the frequency characteristic C1 from the state storage unit 4032 before adjustment.

[0083] Then, the frequency response calculation unit 4033 uses the coefficients ω of one filter. c The transfer function G of mathematical expression 2, defined by τ and δ. F (jω), calculate the input and output gain and the frequency response C2 of the phase lag of a filter.

[0084] Then, the frequency characteristic calculation unit 4033 outputs the frequency characteristic C1 of the input-output gain and phase lag of the servo control unit 100, and the frequency characteristic C2 of the input-output gain and phase lag of the filter, to the state estimation unit 4034.

[0085] The state estimation unit 4034 calculates the estimated values ​​E (E = C1 + C2) of the input / output gain and phase hysteresis frequency characteristics of the servo control unit 100 equipped with one filter by adding the frequency characteristics C1 and C2, and outputs them to the filter setting unit 402. Hereinafter, the estimated value of the input / output gain frequency characteristic in the estimated value E of the input / output gain and phase hysteresis frequency characteristics is set as E1, and the estimated value of the phase hysteresis frequency characteristic is set as E2.

[0086] The filter setting unit 402 sets a filter in the servo control unit 100 based on the estimated value E1 of the frequency characteristics of the input and output gain of the servo control unit 100, thereby enabling it to determine whether the resonance of the selected resonant point has attenuated to an acceptable range.

[0087] The filter setting unit 402 determines whether the resonance of the selected resonant point has attenuated to an acceptable range in the following manner.

[0088] The filter setting unit 402 pre-stores a standard model of the input / output gain of the servo control unit 100. The standard model is a model of the servo control unit with ideal, resonance-free characteristics. For example, the standard model can be based on... Figure 5 The inertia Ja and torque constant K of the model shown t , proportional gain K p Integral gain K I Differential gain K D It is calculated. The inertia Ja is the sum of the inertia of the electric motor and the mechanical inertia.

[0089] Figure 6 This is a characteristic graph showing the frequency characteristics of the input / output gain of the servo control unit in the standard model versus the estimated values ​​of the frequency characteristics of the input / output gain of the servo control unit 100. For example... Figure 6 As shown in the characteristic diagram, the canonical model possesses a frequency domain region A where the input-output gain is above a constant input-output gain, for example, above -20dB, and a frequency domain region B where the input-output gain is less than a constant input-output gain. Figure 6 In region A, the ideal input-output gain of the canonical model is represented by curve MC1 (thick line). Figure 6 In region B, use curve MC 11 (The thick dashed line) represents the ideal virtual input-output gain of the canonical model, where the input-output gain of the canonical model is set to a constant value and represented by the straight line MC. 12 (Thick line) indicates. In Figure 6 In regions A and B, curves RC1 and RC2 represent the estimated value E1 of the input-output gain of the servo control unit, respectively.

[0090] In region A, in the frequency band centered on the selected resonant point, if the curve RC1 of the estimated input-output gain E1 does not exceed the curve MC1 of the ideal input-output gain of the standard model, the filter setting unit 402 determines that the resonance attenuation of the selected resonant point is within the allowable range; if it does, the filter setting unit 402 determines that the resonance attenuation of the selected resonant point is not within the allowable range.

[0091] In the region B above the frequency range where the input-output gain becomes sufficiently small, even if the curve RC1 of the estimated input-output gain E1 exceeds the curve MC of the ideal virtual input-output gain in the canonical model... 11 The impact on stability also decreases. Therefore, in region B, as mentioned above, the input-output gain of the canonical model does not use the ideal gain characteristic curve MC. 11 Instead, it uses a linear MC with a constant input-output gain (e.g., -20dB). 12 The filter setting unit 402 determines the input / output gain curve RC1 within the frequency band centered on the selected resonant point, ensuring that the curve does not exceed the constant input / output gain line MC. 12 In the case of [condition], it is determined that the resonance at the selected resonant point has decayed to the allowable range. If it exceeds the allowable range, there is a possibility of instability. Therefore, it is determined that the resonance at the selected resonant point has not decayed to the allowable range.

[0092] When the resonance attenuation at the selected resonant point is within an acceptable range, the filter setting unit 402 saves the coefficients ω of one filter. c , τ, δ. The filter setting unit 402 changes the coefficients ω of one filter. c The parameters ω, τ, and δ are calculated and output to the frequency response estimation unit 403. By repeating the above process, the filter setting unit 402 obtains the coefficients ω of a filter whose resonant attenuation at the selected resonant point is within the allowable range. c The set of τ and δ.

[0093] The filter setting unit 402 determines the coefficients ω of a filter based on the resonant attenuation to an acceptable range at the selected resonant point. c The cutoff frequencies are calculated by taking the estimated value E2 of the frequency characteristics of the phase lag corresponding to the sets of τ, δ, and so on, and then taking the coefficient ω at the maximum cutoff frequency. c τ and δ determine the coefficients ω of a filter. c τ, δ. The cutoff frequency is, for example, a frequency with a gain characteristic of -3dB or a phase characteristic of -180 degrees, obtained by measuring the frequency response calculated based on the input / output gain of the servo control unit 100. As the cutoff frequency increases, the feedback gain increases, and the response speed increases.

[0094] Next, the filter setting unit 402 selects other resonant points from the k resonant points, and sets the coefficients ω of the other filter in order to attenuate the resonance of the selected other resonant points. c , τ, δ and output to the frequency response estimation unit 403.

[0095] The filter setting unit 402 and the frequency response estimation unit 403 cooperate to sequentially determine the coefficients ω of each filter for the k filters as described above. c The actions of τ and δ determine the coefficients ω of the k filters corresponding to the k resonant points. c τ, δ, are used to initially adjust the coefficients of k filters for each filter.

[0096] The filter setting unit 402 outputs the coefficients of the k initially adjusted filters to the grouping unit 404.

[0097] The above explanation describes the case where the frequency characteristics of input / output gain (amplitude ratio) and phase lag are obtained by operating the servo control unit 100 without filter 130 to detect the resonant point. However, the frequency characteristics of input / output gain and phase lag without filter 130 can also be obtained by other methods. For example, the coefficients ω of filter 130 can be used. c The frequency characteristics of the input-output gain and phase lag of filter 130 are calculated using τ and δ. Furthermore, by activating the servo control unit 100 on which filter 130 is installed, the frequency characteristics of the input-output gain and phase lag are obtained. By subtracting the frequency characteristics of the input-output gain and phase lag of filter 130 from these frequency characteristics, the frequency characteristics of the input-output gain and phase lag without filter 130 can be obtained.

[0098] (Filter grouping)

[0099] Next, the grouping unit 404 will be described to group the initially adjusted k filters and set the operation of m filters (k > m, Nmax ≥ m) with a filter number Nmax or less as a constraint on the number of filters.

[0100] To set the initially adjusted k filters into m filters (k > m), the grouping unit 404 sets a filter combination, which includes one or more filters that combine multiple adjacent filters from the k filters into one filter. Alternatively, the grouping unit 404 sets a filter combination, which includes two or more filter groups that combine three or more adjacent filters from the k filters into two or more filters, which is less than the number of three or more adjacent filters.

[0101] There are four methods, for example, for grouping k filters into m filters.

[0102] The following description illustrates an example of a filter obtained by combining multiple adjacent filters into one filter (hereinafter referred to as a composite filter) in order to set up more than one filter by setting up more than one filter to set up more than one filter (k>m) after initial adjustment of k filters.

[0103] (1) Perform a full search on the combination of filters.

[0104] A composite filter is created by combining multiple adjacent filters into one. The number of filter combinations where k filters are set to m filters becomes the number of (m-1) partitions selected from the (k-1) partitions that divide the k filters. Therefore, it becomes... (k-1) C (m-1) indivual.

[0105] Grouping part 404 is solved (k-1) C (m-1) In a filter combination, the coefficients ω of the composite filter are adjusted one by one. c , τ, δ. The grouping unit 404 sets a filter combination containing a composite filter as filter 130 in the servo control unit 100 for a single filter combination. Furthermore, the grouping unit 404 adjusts the coefficients of the composite filter in such a way that the frequency characteristics of the input / output gain obtained by operating the servo control unit 100 are similar to those obtained by operating the filter setting unit 402, not exceeding the frequency characteristics of the input / output gain used. Figure 5 and Figure 6 The curves illustrating the frequency response of the ideal input-output gain of the canonical model are shown, with the cutoff frequency being the maximum.

[0106] In a filter combination, if there are other composite filters besides the one composite filter mentioned above, the grouping unit 404 adjusts the coefficients of the other composite filters in the same way. A filter combination with the coefficients of one or more composite filters adjusted in this way becomes a grouping candidate.

[0107] When the grouping unit 404 finds one grouping candidate, it finds grouping candidates for other combinations.

[0108] Group 404 repeats the above steps to find... (k-1) C (m-1) Given a group of candidate groups, find the desired result. (k-1) C (m-1) For each group, the objective function value is determined by its respective cutoff frequency. The group that best achieves the objective function value is selected as the optimal combination. If the objective function is the cutoff frequency, the group with the highest cutoff frequency is selected as the optimal combination.

[0109] Figure 7This is a diagram showing the gain characteristics of the input and output gains, the filter settings for multiple resonant points, and an example of grouping. Figure 7 The example shown illustrates generating 10 resonant points, initially adjusting the coefficients of 10 filters in a manner that suppresses these 10 resonant points, and grouping the 10 filters into 5 filters. Figure 7 In the example, group G1 and group G2 are groupings of different filters.

[0110] In the case of a full search of filter combinations Figure 7 The number of filter combinations that group 10 filters into 5 filters is 9C4. Additionally, the number of filter combinations that group 10 filters into 5 or fewer filters is (9C4 + 9C3 + 9C2 + 9C1).

[0111] (2) Searching through machine learning

[0112] In the case of searching using machine learning, such as Figure 8 As shown, the grouping unit 404 is composed of a grouping decision unit 410 and a machine learning unit 420. Figure 8 It is a block diagram representing the structure of a part of a control system centered on a grouping unit.

[0113] In order to set the k filters initially adjusted by the filter setting unit 402 into m filters (k > m), the machine learning unit 420 sets a filter combination with one or more composite filters. This composite filter is obtained by combining multiple adjacent filters from the k filters into one. Alternatively, the filter combination can be set by the grouping determination unit 410, which outputs the filter combination to the machine learning unit 420.

[0114] First, the machine learning unit 420 obtains the coefficients of the k initially adjusted filters from the filter setting unit 402 via the grouping decision unit 410. Then, the machine learning unit 420 sets one or more composite filters, and in a combination of m filters where the k initially adjusted filters are used, a filter combination including one composite filter is set as filter 130 and configured in the servo control unit 100. Next, the machine learning unit 420 obtains the frequency characteristics of the input / output gain obtained by operating the servo control unit 100 from the frequency characteristic measurement unit 300. The machine learning unit 420 ensures that the frequency characteristics of the obtained input / output gain do not exceed the operating frequency of the filter setting unit 402. Figure 5 as well as Figure 6 The ideal input-output gain frequency response curve of the canonical model is described, with the cutoff frequency being the maximum. Machine learning is then performed by adjusting the coefficients ω of a composite filter. cThe grouping determination unit 410 performs the operation of setting a filter combination including a composite filter on the filter 130 and obtaining the frequency characteristics of input-output gain and phase hysteresis from the frequency characteristic measurement unit 300. Alternatively, the machine learning unit 420 can directly set a filter combination including a composite filter on the filter 130 without going through the grouping determination unit 410, and obtain the frequency characteristics of input-output gain and phase hysteresis from the frequency characteristic measurement unit 300.

[0115] In a filter combination where there are other composite filters besides the one mentioned above, the coefficients of those other composite filters are also adjusted in the same way. A filter combination with the coefficients of one or more composite filters adjusted in this way becomes a grouping candidate.

[0116] use Figure 3 This is a specific example illustrating the actions of the Machine Learning Department 420 mentioned above. Here, we will... Figure 3 This diagram is used to illustrate the process of finding one candidate group.

[0117] In the gain characteristics of input and output gain, eight resonant points P1 to P8 are generated. For example, the machine learning unit 420 sets a filter combination with one composite filter for each of the resonant points P2 and P3, P5 and P6, and P7 and P8.

[0118] The machine learning unit 420, for example, sets a filter combination for filter 130 that includes a composite filter set for resonant points P2 and P3. At this time, no composite filter is set for resonant points P5 and P6, or resonant points P7 and P8. The machine learning unit 420 obtains the frequency characteristics of the input / output gain obtained by operating the servo control unit 100, which includes the filter 130 with the composite filter set for resonant points P2 and P3. The machine learning unit 420 ensures that the frequency characteristics of the obtained input / output gain do not exceed the operating frequency of the filter setting unit 402. Figure 5 as well as Figure 6 The ideal input-output gain frequency response curve of the canonical model is described, with the cutoff frequency being the maximum. Machine learning is then used to adjust the coefficients of the composite filter set at resonant points P2 and P3.

[0119] Subsequently, the machine learning unit 420 similarly adjusted the coefficients of the composite filters set for resonant points P5 and P6, and the coefficients of the composite filters set for resonant points P7 and P8. This resulted in a filter combination with adjusted coefficients for the three composite filters set for resonant points P2 and P3, P5 and P6, and P7 and P8, which was then used as a grouping candidate.

[0120] When the machine learning unit 420 finds one grouping candidate, it sets one or more composite filters and finds other grouping candidates by setting the initially adjusted k filters as m filters. Repeating this process, the machine learning unit 420 adjusts the coefficients ω of each filter in the filter combinations of the grouping candidates. c τ and δ are output sequentially to the grouping decision unit 410.

[0121] The details of the structure and operation of the Machine Learning Department 420 will be described later.

[0122] The grouping decision unit 410 sets a filter combination of a grouping candidate obtained by the machine learning unit 420 in the filter 130 to make the servo control unit 100 operate, and uses the frequency characteristics of the phase lag of the servo control unit 100 to obtain the objective function such as the cutoff frequency for a grouping candidate.

[0123] The grouping decision unit 410 calculates the cutoff frequency and other objective functions for the next grouping candidate obtained by the machine learning unit 420 through the same action as the action of calculating the cutoff frequency for one grouping candidate.

[0124] The grouping decision unit 410 repeatedly performs this action to find the filter combination with the best value of the objective function such as the cutoff frequency from multiple grouping candidates obtained by the machine learning unit 420.

[0125] (3) Search based on the determined rules

[0126] Based on the determined rules, the grouping unit 404 sets the initially adjusted k filters to m filters (k > m). Rules may include, for example, not grouping filters with low damping (attenuation ratio) (i.e., grouping weak filters), starting with high-frequency filters, and not grouping filters whose attenuation center frequency is far from the filter's.

[0127] The method of obtaining a composite filter that combines multiple filters into one is to perform the combination of the filters described above (1) by the same action as the action of finding group candidates described in the full search.

[0128] When multiple candidate groups are obtained based on this rule, the grouping unit 404 selects the filter combination with the best value of objective function such as cutoff frequency from the obtained candidate groups.

[0129] (4) Summarize filters one by one

[0130] Grouping unit 404 aggregates the initially adjusted k filters one by one, setting them as m filters (k > m). For example, by reducing the k filters by one to aggregate (k-1) filters, reducing the (k-1) filters by one to aggregate (k-2) filters, and so on until there are m filters. Evaluation is not required in this case.(k-1) C (m-1) The pattern.

[0131] The method for obtaining a composite filter by combining multiple filters into one is as follows.

[0132] Grouping section 404 adjusts the coefficient ω of one composite filter. c τ, δ, such that the frequency response of the input-output gain obtained by operating the servo control unit 100, which sets (k-1) filters, including one aggregated adjacent filter, as filter 130, is similar to the operation of the filter setting unit 402, not exceeding the frequency response of the input-output gain obtained by operating the servo control unit 100. Figure 5 as well as Figure 6 The curves illustrating the frequency response of the ideal input-output gain of the canonical model are shown, with the cutoff frequency being the maximum.

[0133] Next, the grouping unit 404 adjusts the coefficients of the newly set composite filter so that the frequency characteristics of the input-output gain obtained by setting a composite filter that combines adjacent filters from (k-1) filters to form (k-2) filters, and setting (k-2) filters as filter 130, are the same as those of the filter setting unit 402, not exceeding the frequency characteristics of the input-output gain. Figure 5 as well as Figure 6 The curves illustrating the ideal input-output gain frequency response of the canonical model are given, with the cutoff frequency being maximized.

[0134] Repeat these steps to find m filter combinations.

[0135] <Machine Learning Department>

[0136] The structure and operation of the machine learning unit 420 when the grouping unit 404 obtains the grouping candidates of the filter 130 through the machine learning described in (2) above will be explained below.

[0137] The following explanation describes the reinforcement learning process performed by the Machine Learning Unit 420. However, machine learning is not specifically limited to reinforcement learning; for example, the Machine Learning Unit 420 can also perform supervised learning.

[0138] In order to set the initially adjusted k filters as m filters (k>m) in the filter setting unit 402, the machine learning unit 420 sets a filter combination with one or more composite filters, which is obtained by combining multiple adjacent filters from the k filters into one.

[0139] The machine learning unit 420 sets one or more composite filters. In a combination of m filters after initial adjustment (k filters), a filter combination including one composite filter is set as filter 130 and configured in the servo control unit 100. The machine learning unit 420 obtains the frequency characteristics of input / output gain and phase hysteresis obtained from the frequency characteristic measurement unit 300 by operating the servo control unit 100, as state S. Then, the machine learning unit 420 calculates the coefficients ω of the composite filter of filter 130 in the servo control unit 100 related to state S. c The adjustment of the values ​​of τ and δ is considered as Q-learning for behavior A. In a filter combination, if other composite filters exist besides the one mentioned above, Q-learning is also performed on the coefficients of these other composite filters. As is well known to those skilled in the art, the purpose of Q-learning is to select the behavior A with the highest value Q(S, A) from the desirable behaviors A at a certain state S as the optimal behavior.

[0140] Specifically, an intelligent agent (machine learning device) selects various actions A in a certain state S. For the action A at this time, it selects a better action based on the reward given, thereby learning the correct value Q(S, A).

[0141] Furthermore, since the goal is to maximize the total future returns, the final result should be Q(S, A) = E[Σ(γ t )r t The objective is E[]. Here, E[] represents the expected value, t is time, γ is a parameter referred to as the discount rate (described later), and r t Let Σ be the reward at time t, and Σ be the sum based on time t. The expected value in this formula is the expected value under the condition of state change according to optimal behavior. Such an updated formula for value Q(S, A) can be expressed, for example, by the following mathematical formula 3 (hereinafter referred to as formula 3).

[0142] [Formula 3]

[0143]

[0144] In the above mathematical formula 3, S t A represents the state of the environment at time t. t Represents the behavior at time t. Through behavior A... t The state change is S t+1 r t+1 This represents the reward obtained through the change of this state. Furthermore, the term with `max` is in state S. t+1The value obtained by multiplying the Q-value of the selected action with the highest known Q-value by γ is the result of taking action A as the first choice. Here, γ is a parameter of 0 < γ ≤ 1, referred to as the discount rate. Additionally, α is the learning coefficient, set to the range of 0 < α ≤ 1.

[0145] The above mathematical expression 3 represents the result of trial A. t The result and the return r t+1 Update status S t The following behavior A t Value Q(S) t A t The method.

[0146] The machine learning unit 420 observes the state information S of the frequency characteristics at each frequency, including the input-output gain and phase hysteresis, output from the frequency characteristic measurement unit 300, and determines action A. The machine learning unit 420 returns a report each time action A is performed. The report will be described later.

[0147] In Q-learning, the machine learning unit 420, for example, searches on a trial-and-error basis for the optimal action A that maximizes the sum of future rewards. Thus, the machine learning unit 420 can select the optimal action A for state S (the optimal coefficients ω of a composite filter). c 、τ、δ).

[0148] Figure 9 This is a block diagram illustrating a machine learning unit 420 according to an embodiment of the present invention.

[0149] In order to perform the above reinforcement learning, such as Figure 9 As shown, the machine learning unit 420 includes a state information acquisition unit 421, a learning unit 422, a behavior information output unit 423, a value function storage unit 424, and a grouping candidate output unit 425.

[0150] The status information acquisition unit 421 obtains the coefficients ω of the k filters after initial adjustment from the filter setting unit 402 via the grouping decision unit 410. c τ, δ. Additionally, the state information acquisition unit 421, in order to set the initially adjusted k filters as m filters (k > m), sets a filter combination with one or more composite filters. This composite filter is obtained by combining multiple adjacent filters from the k filters into one. The user pre-generates the coefficients ω of the composite filter at the initial Q-learning time point. c The initial values ​​of ω, τ, and δ. The coefficients of the filter other than the composite filter coefficients are values ​​after initial adjustment. In this embodiment, reinforcement learning is used to adjust the coefficients ω of the user-generated composite filter. c The initial settings of τ and δ are adjusted to their optimal values.

[0151] At the initial point when Q-learning begins, the state information acquisition unit 421 outputs to the behavior information generation unit 4223 information on the multiple adjacent filters that should be removed from the k filters after initial adjustment, and the initial values ​​of the coefficients of the permuted composite filter.

[0152] Additionally, the status information acquisition unit 421 obtains the coefficients ω of the filter 130, which is configured to include a filter combination of one composite filter. c The states S, including input / output gain (amplitude ratio) and phase hysteresis, are obtained from the frequency response measurement unit 300 via the grouping determination unit 410 by driving the servo control unit 100 using a speed command (sine wave), and are output to the learning unit 422. This state information S corresponds to the environmental state S in Q learning.

[0153] The learning unit 422 is responsible for learning the value Q(S, A) when choosing a certain action A in a certain environmental state S. The learning unit 422 includes a reward output unit 4221, a value function update unit 4222, and an action information generation unit 4223.

[0154] The reward output unit 4221 is the part that calculates the reward when behavior A is selected in a certain state S.

[0155] First, the servo control unit 100 of the filter 130, with coefficients of the composite filter set to initial values, operates. The frequency characteristics of the input / output gain output from the frequency characteristic measurement unit 300, including the input / output gain and phase hysteresis frequency characteristics, exceed those of the filter setting unit 402, similarly exceeding the operating frequency characteristics of the filter. Figure 5 as well as Figure 6 The ideal input-output gain frequency response curve of the canonical model is given a negative return, and the coefficients of the composite filter are corrected according to the initial values.

[0156] On the other hand, when the input / output gain is below the value of the input / output gain in the normalized model, the feedback output unit 4221 uses the frequency characteristic of the phase lag output from the frequency characteristic measurement unit 300 to determine the first cutoff frequency that becomes the value of the first objective function. The cutoff frequency is, for example, a frequency with a gain characteristic of -3dB obtained from the Bode plot of the frequency response calculated based on the input / output gain of the servo control unit 100, or a frequency with a phase characteristic of -180 degrees. As the cutoff frequency increases, the feedback gain increases, and the response speed becomes faster.

[0157] Next, the servo control unit 100 of the filter 130, which has adjusted the initial values ​​of the coefficients of the composite filter, operates. The frequency characteristics of the input / output gain output from the frequency characteristic measurement unit 300, including the input / output gain and phase hysteresis frequency characteristics, exceed those of the filter setting unit 402, just as the operation of the filter setting unit 402 has. Figure 5 as well as Figure 6 The ideal input-output gain frequency response curve of the canonical model is given a negative return, which further modifies the coefficients of the composite filter.

[0158] On the other hand, when the input-output gain is below the value of the input-output gain of the normalized model, the return output unit 4221 uses the frequency characteristic of the phase lag output from the frequency characteristic measurement unit 300 to determine the second cutoff frequency that becomes the value of the second objective function.

[0159] Then, the reward output unit 4221 calculates the difference between the first cutoff frequency and the second cutoff frequency, which constitute the evaluation function. Furthermore, the reward output unit 4221 assigns a positive reward when the second cutoff frequency is greater than the first cutoff frequency, assigns a zero reward when the second cutoff frequency is the same as the first cutoff frequency, and assigns a negative reward when the second cutoff frequency is less than the first cutoff frequency.

[0160] Next, the servo control unit 100, which has further adjusted coefficients of the composite filter, operates in the same way as the operation to determine the second cutoff frequency. If the input-output gain is below the value of the input-output gain of the standard model, the feedback output unit 4221 uses the frequency characteristic of the phase lag output from the frequency characteristic measurement unit 300 to determine the third cutoff frequency, which becomes the value of the third objective function.

[0161] Then, the reward output unit 4221 calculates the difference between the second cutoff frequency and the third cutoff frequency, which become the evaluation function. The reward output unit 4221 then assigns a positive reward when the third cutoff frequency is greater than the second cutoff frequency, a zero reward when the third cutoff frequency is the same as the second cutoff frequency, and a negative reward when the third cutoff frequency is less than the second cutoff frequency.

[0162] Furthermore, when the input-output gain described above is below the value of the input-output gain of the standard model, the return output unit 4221 repeatedly performs the same operation as the operation of giving a return using the second cutoff frequency and the third cutoff frequency by adjusting the coefficients of the composite filter.

[0163] The above explains the return output unit 4221.

[0164] The value function update unit 4222 learns Q based on the state S, behavior A, the state S′ when behavior A is applied to state S, and the reward obtained as described above, thereby updating the value function Q stored in the value function storage unit 424.

[0165] The value function Q can be updated through online learning, batch learning, or mini-batch learning.

[0166] Online learning is a learning method that applies an action A to the current state S, and immediately updates the value function Q whenever S transitions to a new state S′. Batch learning, on the other hand, involves repeatedly applying action A to the current state S to transform it into a new state S′, thereby collecting learning data, and using all the collected learning data to update the value function Q. Mini-batch learning, however, is an intermediate method between online and batch learning, where the value function Q is updated only after a certain amount of learning data has accumulated.

[0167] The behavior information generation unit 4223 selects behavior A during the Q-learning process for the current state S. During Q-learning, the behavior information generation unit 4223 adjusts the coefficients ω of the composite filter for filter 130. c The action of τ (equivalent to behavior A in Q learning) generates behavior information A and outputs the generated behavior information A to the behavior information output unit 423.

[0168] Alternatively, the behavior information generation unit 4223 may also adopt the following countermeasures: select behavior A′ by a greedy algorithm that selects the behavior A′ with the highest value Q(S,A) among the currently estimated values ​​of behavior A, or by an ε-greedy algorithm that randomly selects behavior A′ with a small probability ε, and otherwise selects the behavior A′ with the highest value Q(S,A).

[0169] The behavior information output unit 423 is the part that sends the behavior information A output from the learning unit 422 to the servo control unit 100. As described above, based on this behavior information, the coefficients ω of the composite filter of the current state S, i.e., the currently set filter 130, are adjusted. c , τ, δ, thus transitioning to the next state S′.

[0170] The value function storage unit 424 is a storage device for storing the value function Q. The value function Q can also be stored as a table (hereinafter referred to as the behavior value table) for each state S and behavior A. The value function Q stored in the value function storage unit 424 is updated by the value function update unit 4222. Furthermore, the value function Q stored in the value function storage unit 424 can also be shared among other machine learning units 420. If the value function Q is shared among multiple machine learning units 420, reinforcement learning can be performed distributed across each machine learning unit 420, thus improving the efficiency of reinforcement learning.

[0171] The above describes reinforcement learning in the case where one composite filter is set in filter 130. However, if there are other composite filters in filter 130 besides one composite filter, reinforcement learning is also performed on the coefficients of the other composite filters in the same way.

[0172] Furthermore, reinforcement learning is also performed when setting more than one composite filter, and there are other combinations of setting the initially adjusted k filters as m filters.

[0173] The grouped candidate output unit 425 calculates the coefficients ω of each filter in a filter combination containing one or more composite filters with the largest value Q(S,A) based on the updated value function Q obtained by the value function update unit 4222 through Q-learning. c The coefficients of the filter combination, τ and δ, are used as grouping candidates and output to the grouping decision unit 410.

[0174] More specifically, the group candidate output unit 425 acquires the value function Q stored in the value function storage unit 424. As described above, this value function Q is a function updated by Q-learning through the value function update unit 4222. Then, based on the value function Q, the group candidate output unit 425 calculates the coefficients ω of each filter in a filter combination that includes one or more composite filters with the largest value Q(S, A). c The coefficients of the filter combination, τ and δ, are used as grouping candidates and output to the grouping decision unit 410.

[0175] By setting up one or more composite filters, reinforcement learning is also applied to other combinations of m filters, using the initially adjusted k filters as the basis for further processing. The coefficients ω of each filter in each combination of filters containing the composite filter with the highest value Q(S,A) are then determined. c τ, δ. The group candidate output unit 425 outputs the coefficients of the filter combination as the next group candidate to the group decision unit 410. For example, when Figure 7 When group G1 is a filter combination, group G2 becomes an example of other filter combinations.

[0176] As described above, the machine learning unit 420 learns the coefficients ω of the composite filter. c Find the optimal values ​​of τ and δ to determine the candidate groups.

[0177] The functional blocks included in the control system 10 have been described above.

[0178] To implement these functional blocks, the control system 10, servo control unit 100, or control auxiliary unit 400 includes an arithmetic processing unit such as a CPU (Central Processing Unit). Additionally, the control system 10, servo control unit 100, or control auxiliary unit 400 also includes auxiliary storage devices such as HDDs (Hard Disk Drives) that store various control programs such as application software or operating systems (OS), and main storage devices such as RAMs (Random Access Memory) that store data temporarily required by the arithmetic processing unit when executing programs.

[0179] Then, in the control system 10, the servo control unit 100, or the control auxiliary unit 400, the arithmetic processing unit reads application software or operating system from the auxiliary storage device, expands the read application software or operating system in the main storage device, and performs arithmetic processing based on these application software or operating system. Furthermore, the arithmetic processing unit controls various hardware components of each device based on the results of this arithmetic processing. Thus, the functional blocks of this embodiment are implemented. That is, this embodiment can be implemented through hardware and software cooperation.

[0180] When the computational load of the control auxiliary unit 400 is high, for example, by equipping a personal computer with a GPU (Graphics Processing Unit), a technology called GPGPU (General-Purpose Computing on Graphics Processing Units) can be used to perform high-speed processing. Furthermore, for even faster processing, multiple computers equipped with such GPUs can be used to build a computer cluster, and parallel processing can be performed by the multiple computers in the computer cluster.

[0181] Next, a flowchart will be used to explain the operation of the control auxiliary unit 400. Figure 10 This is a flowchart showing the actions of the control auxiliary unit.

[0182] In step S11, the resonance detection unit 401 obtains the frequency characteristics of the input-output gain (amplitude ratio) of the servo control unit 100 from the frequency characteristic measurement unit 300 when the servo control unit 100 operates without the filter 130.

[0183] In step S12, the resonance detection unit 401 detects k resonant points of the frequency characteristics (gain characteristics) of the input and output gain.

[0184] In step S13, the filter setting unit 402, in cooperation with the frequency response estimation unit 403, sets up k filters corresponding to k resonant points, and performs initial adjustments to the coefficients of the k filters according to each filter.

[0185] The number of filters initially adjusted, n ini Let the number be the same as the number of resonant points (k = n) ini However, it is also possible to compare the number of resonant points k with the number n smaller. ini (k>n ini The filter is initially adjusted.

[0186] In step S14, the grouping unit 404 groups the initially adjusted k filters and sets m (k>m) filters.

[0187] In step S15, the control auxiliary unit 400 determines whether to continue the process of grouping k filters and setting m (k>m) filters. If it continues, it returns to step S14; if it does not continue, it ends the operation of the control auxiliary unit.

[0188] According to the implementation method described above, resonance suppression can be achieved, and the allocation of filters can be determined in a way that satisfies the constraint of the number of filters.

[0189] (Second Implementation)

[0190] In the first embodiment, the frequency response measurement unit 300 calculates the frequency response based on a frequency-changing sinusoidal signal, i.e., speed command and speed feedback, when measuring the frequency response of the input-output gain (amplitude ratio) and phase hysteresis of the servo control unit 100. In this embodiment, the frequency generation unit 200 inputs a sinusoidal signal to the preamplifier of the current control unit 140 while changing the frequency. Then, the frequency response measurement unit 300 calculates the frequency response based on the sinusoidal signal input to the preamplifier of the current control unit 140 and the output of the speed control unit 120, while measuring the frequency response of the input-output gain and phase hysteresis of the servo control unit 100.

[0191] Figure 11 This is a block diagram illustrating the control system according to the second embodiment of this disclosure. Figure 11 In the middle, to and Figure 1 The components of the control system 10 shown are labeled with the same reference numerals and their descriptions are omitted.

[0192] like Figure 11 As shown, the control system 10A has an adder 160 installed before the subtractor 170. A sinusoidal signal with varying frequency output from the frequency generation unit 200 is input to the adder 160. The adder 160 is connected to the subtractor 170, and the current control unit 140 is connected to the amplifier 180. The amplifier 180 has a current detector, and the current detected by the current detector is input to the subtractor 170. The subtractor 170, the current control unit 140, and the amplifier 180 constitute a current feedback loop, which is included in the speed feedback loop. The sinusoidal signal corresponds to the first signal with varying frequency, and the output of the filter 130 corresponds to the second signal input to the current feedback loop in the speed feedback loop.

[0193] The inductance of the motor 150 changes non-linearly depending on the current flowing through it due to factors such as magnetic saturation. When the servo parameters change from those before adjustment to those after adjustment, the torque command input to the current control unit 140 changes, and even with a constant current gain in the current control unit 140, the current flowing through the motor 150 also changes. As the current flowing through the motor 150 changes, and the inductance changes non-linearly, the characteristics of the current feedback loop also change non-linearly.

[0194] In this embodiment, the input signal level to the subtractor 110 is set to zero. The frequency generation unit 200 inputs a sine wave signal to the preamp of the current control unit 140 while changing the frequency. The frequency response measurement unit 300 measures the frequency characteristics of the input-output gain and phase hysteresis of the servo control unit 100 based on the sine wave signal and the output of the speed control unit 120. As a result, the input to the current feedback loop is constant, thus maintaining the linearity of the current feedback loop characteristics while obtaining multiple resonances through the control auxiliary unit 400.

[0195] The first and second embodiments have been described above, but in the first and second embodiments, the filter setting unit 402 may also include the same... Figure 9 The machine learning unit 420 shown has the same structure as the machine learning unit, and uses the machine learning unit to perform initial adjustments on the coefficients of k filters for each filter. The initial adjustment of each filter is related to adjusting the coefficients ω to the optimal values ​​for the composite filter. c Since the actions of τ and δ are the same, their descriptions are omitted.

[0196] In the embodiments described above, the objective function is not limited to the cutoff frequency; examples include |1 - (closed-loop gain characteristic)|. 2 Or |1-(closed-loop transfer function)| 2 The closed-loop transfer function can be derived from the gain A(ω) and phase lag θ(ω) of the Bode plot using G(jω) = A(ω) × e -jθ(ω) To calculate. Here, closed loop refers to the speed feedback loop consisting of subtractor 110, speed control unit 120, filter 130, current control unit 140, and motor 150.

[0197] (Modified Example)

[0198] In addition to variations of the control system Figure 11 In addition to its structure, it also has the following structures.

[0199] <Example of a control auxiliary unit connecting to a servo control unit via a network>

[0200] Figure 12 This is a block diagram representing a variation of the control system. Figure 12 The control system 10B shown can be applied to Figure 1 as well as Figure 11 The control systems 10 and 10A of the first and second embodiments are shown. The control system 10B differs from control systems 10 and 10A in that n (n is a natural number of 2 or more) servo control units 100-1 to 100-n are connected to n control auxiliary units 400-1 to 400-n via a network 500, and each includes a frequency generation unit 200 and a frequency characteristic measurement unit 300. The control auxiliary units 400-1 to 400-n have... Figure 1 The control auxiliary unit 400 shown has the same structure. Servo control units 100-1 to 100-n correspond to servo control devices, and control auxiliary units 400-1 to 400-n correspond to control auxiliary devices. Furthermore, one or both of the frequency generation unit 200 and the frequency characteristic measurement unit 300 may, of course, be provided outside the servo control units 100-1 to 100-n.

[0201] Here, the servo control unit 100-1 and the control auxiliary unit 400-1 are connected in a one-to-one manner, enabling communication. The servo control units 100-2 to 100-n and the control auxiliary units 400-2 to 400-n are also connected to the servo control unit 100-1 and the control auxiliary unit 400-1 in the same way. Figure 12In this configuration, n groups of servo control units 100-1 to 100-n and control auxiliary units 400-1 to 400-n are connected via network 500. However, the n groups of servo control units 100-1 to 100-n and control auxiliary units 400-1 to 400-n can also be directly connected via a connection interface. For example, multiple groups of these servo control units 100-1 to 100-n and control auxiliary units 400-1 to 400-n can be installed in the same factory, or they can be installed in different factories.

[0202] Furthermore, Network 500 can be, for example, a LAN (Local Area Network), the Internet, the public telephone network, or a combination thereof, built within a factory. There are no particular limitations regarding the specific communication methods used in Network 500, or whether the connection is wired or wireless.

[0203] <Degrees of freedom of system structure>

[0204] In the above embodiments, the servo control units 100-1 to 100-n and the control assistance units 400-1 to 400-n are respectively set as a 1-to-1 group and connected in a communicable manner. However, for example, one control assistance unit may be connected to multiple servo control units via network 500 in a communicable manner to implement control assistance for each servo control unit.

[0205] At this point, the functions of a single control and auxiliary unit can also be configured as a distributed processing system appropriately distributed across multiple servers. Alternatively, the functions of a single control and auxiliary unit can also utilize virtual server functionality in the cloud.

[0206] Furthermore, if there are n control auxiliary units 400-1 to 400-n corresponding to n servo control units 100-1 to 100-n of the same type, name, specifications, or series, the estimation results of each control auxiliary unit 400-1 to 400-n can be shared. This allows for the construction of a better model.

[0207] The first and second embodiments and their variations have been described above. Each component included in the control system of each embodiment and variation can be implemented using hardware, software, or a combination thereof. Furthermore, the servo control method that operates through the cooperation of each component included in the aforementioned control system can also be implemented using hardware, software, or a combination thereof. Here, implementation using software means implementing it by loading and executing a program into a computer.

[0208] Programs can be stored and provided to a computer using various types of non-transitory computer-readable media. Non-transitory computer-readable media include various types of tangible storage media. Examples of non-transitory computer-readable media include magnetic recording media (e.g., hard disk drives), optical-magnetic recording media (e.g., optical discs), CD-ROMs (Read Only Memory), CD-Rs, CD-R / Ws, semiconductor memories (e.g., mask ROMs, PROMs (Programmable ROMs), EPROMs (Erasable PROMs), flash memory ROMs, and RAMs (random access memory).

[0209] The above-described embodiments are preferred embodiments of the present invention, but the scope of the present invention is not limited to the above-described embodiments. Various modifications can be made without departing from the spirit of the present invention.

[0210] The control auxiliary device, control system, and filter adjustment method disclosed herein include the embodiments described above, and can be implemented in various ways having the following structures.

[0211] (1) A control assist device (e.g., a control assist unit 400) that assists in setting a plurality of first filters (e.g., filters 130-1 to 130-m), the plurality of first filters being disposed in a servo control device (e.g., a servo control unit 100) for controlling a motor (e.g., a motor 150), wherein the control assist device comprises:

[0212] The resonance detection unit (e.g., resonance detection unit 401) detects multiple resonance points in the frequency characteristics of the input-output gain of the servo control device, which are determined based on the input and output signals that change in frequency.

[0213] A filter setting unit (e.g., filter setting unit 402) sets a plurality of second filters, exceeding the number of the plurality of first filters, in order to suppress the plurality of resonant points; and

[0214] A grouping unit (e.g., grouping unit 404) groups the plurality of second filters to set the plurality of first filters.

[0215] According to this control auxiliary device, resonance suppression can be achieved, and the allocation of filters can be determined in a way that satisfies the constraint of the number of filters.

[0216] (2) The control aid device according to (1) above, wherein the grouping unit comprises: a machine learning unit (e.g., machine learning unit 420) that calculates grouping candidates, wherein the grouping candidates are provided with at least one composite filter that combines two or more of the plurality of second filters; and a grouping decision unit (e.g., grouping decision unit 410) that calculates a combination of the plurality of first filters based on the plurality of grouping candidates calculated by the machine learning unit.

[0217] (3) According to the control auxiliary device described in (1) above, the grouping unit sets at least one composite filter that combines two or more filters of the plurality of second filters based on a predetermined rule, thereby determining the combination of the plurality of first filters.

[0218] (4) According to the control auxiliary device described in (1) above, the grouping unit summarizes the plurality of second filters one by one to obtain the combination of the plurality of first filters.

[0219] (5) According to the control auxiliary device described in (1) above, the grouping unit shall be provided with at least one composite filter that combines two or more of the plurality of second filters, and shall search for all modes for setting the plurality of first filters, and determine the combination of the plurality of first filters from the all modes.

[0220] (6) In any one of (1) to (5) above, the filter in the plurality of first filters is a notch filter and / or a low-pass filter.

[0221] (7) In the control auxiliary device according to any one of (1) to (6) above, when the number of the plurality of second filters is less than a certain number, the grouping unit also groups the plurality of second filters to set the plurality of first filters.

[0222] (8) A control system comprising:

[0223] A servo control device (e.g., servo control unit 100) that controls a motor (e.g., motor 150); and

[0224] The control auxiliary device (e.g., control auxiliary unit 400) described in any one of (1) to (7) above detects multiple resonant points in the frequency characteristics of the input-output gain and input-output phase lag of the servo control device, and sets multiple second filters exceeding the number of multiple first filters set in the servo control device in order to suppress the multiple resonant points, and groups the multiple second filters to set the multiple first filters.

[0225] According to this control system, the allocation of filters can be determined in a way that satisfies the constraint of the number of filters while achieving resonance suppression.

[0226] (9) The control system according to (8) above comprises: a frequency generating device that generates a frequency-changing signal and inputs the signal to the servo control device; and

[0227] A frequency characteristic measuring device, which measures the frequency characteristics of the input-output gain and phase hysteresis of the servo control device based on the signal and the output signal of the servo control device.

[0228] (10) According to the control system described in (8) or (9) above, the servo control device comprises: a current feedback loop that controls the current flowing through the motor; and a feedback loop that includes the current feedback loop and has the filter.

[0229] The control system includes:

[0230] A frequency generation device that generates a first signal with a changing frequency and inputs the first signal into the current feedback loop; and

[0231] The frequency response measurement unit measures the frequency characteristics of the input-output gain and phase hysteresis of the servo control device based on the first signal and the second signal input to the current feedback loop in the feedback loop.

[0232] (11) A filter adjustment method is a filter adjustment assistance method for a control assistance device (e.g., a control assistance unit 400), wherein the control assistance device assists in setting a plurality of first filters, the first filters being provided in a servo control device (e.g., a servo control unit 100) that controls a motor (e.g., a motor 150), wherein the filter adjustment method performs the following processing:

[0233] Multiple resonant points in the frequency characteristics of the input-output gain of the servo control device, determined based on the input and output signals with frequency variations;

[0234] To suppress the plurality of resonant points, a plurality of second filters are configured, exceeding the number of the plurality of first filters; and

[0235] The plurality of second filters are grouped to set the plurality of first filters.

[0236] According to this filter adjustment method, the allocation of filters can be determined in a way that satisfies the constraint of the number of filters while achieving resonance suppression.

[0237] Symbol Explanation

[0238] 10, 10A, 10B control systems;

[0239] Servo control units 100, 100-1 to 100-n;

[0240] 110 subtractor;

[0241] 120 speed control unit;

[0242] 130, 130-1 to 130-m filters;

[0243] 140 Current Control Unit;

[0244] 150 electric motor;

[0245] 200 frequency generation unit;

[0246] 300 Frequency Response Measurement Unit;

[0247] 400, 400-1 to 400-n control auxiliary units;

[0248] 401 Resonance Detection Unit;

[0249] 402 Filter Setting Section;

[0250] 403 Frequency Response Estimation Section;

[0251] Group 404;

[0252] Group 410 Decision-Making Department;

[0253] 420 Machine Learning Department;

[0254] 421 Status Information Acquisition Department;

[0255] 422 Study Department;

[0256] 423 Behavioral Information Output Department;

[0257] 424 Value Function Storage Unit;

[0258] 425 group candidate output section;

[0259] 500 network.

Claims

1. A control auxiliary device for assisting in setting a plurality of first filters, said plurality of first filters being disposed in a servo control device for controlling a motor, characterized in that, The control auxiliary device includes: The resonance detection unit detects multiple resonant points in the frequency characteristics of the input-output gain of the servo control device, which are determined based on the input and output signals that change in frequency. The filter setting unit sets a plurality of second filters in order to suppress the plurality of resonant points, which exceeds the number of the plurality of first filters; as well as The grouping unit groups the plurality of second filters to configure the plurality of first filters. The grouping unit combines two or more filters from the plurality of second filters to set at least one composite filter, thereby setting the plurality of first filters.

2. The control auxiliary device according to claim 1, characterized in that, The grouping unit has: The machine learning unit determines grouping candidates, each grouping candidate having at least one of the aforementioned composite filters; and The grouping decision unit determines the combination of the plurality of first filters based on the plurality of grouping candidates obtained by the machine learning unit.

3. The control auxiliary device according to claim 1, characterized in that, The grouping unit sets at least one of the composite filters based on a predetermined rule, thereby determining the combination of the plurality of first filters.

4. The control auxiliary device according to claim 1, characterized in that, The grouping unit summarizes the plurality of second filters one by one to obtain the combination of the plurality of first filters.

5. The control auxiliary device according to claim 1, characterized in that, The grouping unit is provided with at least one of the composite filters, searches for all modes for setting the plurality of first filters, and obtains the combination of the plurality of first filters from the plurality of modes.

6. The control auxiliary device according to any one of claims 1 to 5, characterized in that, The filters among the plurality of first filters are notch filters and / or low-pass filters.

7. The control auxiliary device according to any one of claims 1 to 5, characterized in that, When the number of the plurality of second filters is less than a certain number, the grouping unit also groups the plurality of second filters to set the plurality of first filters.

8. A control system, characterized in that, have: Servo control device that controls the motor; and The control auxiliary device according to any one of claims 1 to 7 detects multiple resonant points in the frequency characteristics of the input-output gain and the phase lag of the input-output of the servo control device, and sets multiple second filters exceeding the number of multiple first filters set in the servo control device in order to suppress the multiple resonant points, and groups the multiple second filters to set the multiple first filters.

9. The control system according to claim 8, characterized in that, The control system includes: A frequency generating device that generates a frequency-varying signal and inputs the signal into the servo control device; and A frequency characteristic measuring device, which measures the frequency characteristics of the input-output gain and phase hysteresis of the servo control device based on the signal and the output signal of the servo control device.

10. The control system according to claim 8 or 9, characterized in that, The servo control device includes: a current feedback loop that controls the current flowing through the motor; and a feedback loop that includes the current feedback loop and has the filter. The control system includes: A frequency generation device that generates a first signal with a changing frequency and inputs the first signal into the current feedback loop; and The frequency response measurement unit measures the frequency characteristics of the input-output gain and phase hysteresis of the servo control device based on the first signal and the second signal input to the current feedback loop in the feedback loop.

11. A filter adjustment method, which is a filter adjustment auxiliary method for a control auxiliary device, wherein the control auxiliary device assists in setting a plurality of first filters, the plurality of first filters being disposed in a servo control device for controlling a motor, characterized in that, The filter adjustment method performs the following processing: Multiple resonant points in the frequency characteristics of the input-output gain of the servo control device, determined based on the input and output signals with frequency variations; In order to suppress the multiple resonant points, a plurality of second filters are set in a number exceeding that of the plurality of first filters; The plurality of second filters are grouped to set the plurality of first filters. The plurality of first filters are configured by combining two or more of the plurality of second filters to form at least one composite filter.

Images