Control support device, control system, and filter adjustment support method
The control support device and method efficiently distribute filters beyond the number restriction to suppress resonance points, addressing the limitation of servo control devices and improving system stability.
Patent Information
- Application Number
- JP2022551934
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2020-09-23
- Filing Date
- 2021-09-16
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2041-09-16
AI Technical Summary
Existing servo control devices face limitations in setting multiple filters to suppress multiple resonance points, as the number of filters is restricted, making it difficult to determine their optimal distribution for effective resonance suppression.
A control support device and method that includes a resonance detection unit to identify resonance points, a filter setting unit to exceed the filter limit, and a grouping unit to set filters efficiently, allowing for the distribution of filters to satisfy the restriction while achieving resonance suppression.
The solution enables the determination of filter distribution to satisfy the number restriction while effectively suppressing resonance, enhancing the control system's stability and performance.
Smart Images

Figure 0007586920000004 
Figure 0007586920000005 
Figure 0007586920000006
Abstract
Description
[Technical field]
[0001] The present invention relates to a control assistance device for providing assistance in setting a plurality of filters provided in a servo control device that controls a motor, a control system including the control assistance device and the servo control device, and a filter adjustment method. [Background technology]
[0002] In a machine having multiple resonance points, when attempting to improve the feedback characteristics while maintaining stability, it is necessary to suppress the multiple resonance points with multiple filters. For example, Patent Document 1 describes a control system equipped with a servo control device that suppresses multiple resonance points in a machine having multiple resonance points using multiple filters, and a machine learning device that optimizes the filter coefficients. Patent Document 1 describes a control system in which, when a machine has multiple resonance points, multiple filters are provided in a servo control unit (servo control device) corresponding to each resonance point and connected in series to attenuate all resonances. Patent Document 1 also describes that a machine learning device uses machine learning to sequentially determine optimal values for attenuating the resonance points for the coefficients of multiple filters. [Prior art documents] [Patent documents]
[0003] [Patent Document 1] JP 2020-57211 A Summary of the Invention [Problem to be solved by the invention]
[0004] As described above, in a machine having a plurality of resonance points, it is required to suppress the plurality of resonance points by a plurality of filters, but the number of filters that can be set in a servo control device is limited. In order to suppress multiple resonance points, it is important to determine how to allocate a finite number of filters, but it is difficult to use machine learning or optimization algorithms to haphazardly search for and determine filter allocations while suppressing resonances. Therefore, there is a need for a control assistance device, a control system, and a filter adjustment method that can determine filter allocation so as to satisfy the constraint on the number of filters while achieving resonance suppression. [Means for solving the problem]
[0005] (1) A first aspect of the present disclosure is a control assistance device that provides assistance for setting a first plurality of filters provided in a servo control device that controls a motor, the control assistance device comprising: a resonance detection unit that detects a plurality of resonance points in a frequency characteristic of an input / output gain of the servo control device, the frequency characteristic being measured based on an input signal and an output signal whose frequency changes; a filter setting unit that sets a second plurality of filters, the number of which exceeds the first plurality of filters, in order to suppress the plurality of resonance points; a grouping unit that groups the second plurality of filters to set the first plurality of filters; It is a control assistance device equipped with the above.
[0006] (2) A second aspect of the present disclosure provides a servo control device that controls a motor, The control assistance device according to the above (1), which detects a plurality of resonance points in a frequency characteristic of an input / output gain and an input / output phase delay of the servo control device, and sets a number of second filters that exceeds a number of first filters set in the servo control device in order to suppress the plurality of resonance points, and sets the first filters by grouping the second filters; It is a control system equipped with the above.
[0007] (3) A third aspect of the present disclosure is a filter adjustment support method for a control support device that provides support for setting a first plurality of filters provided in a servo control device that controls a motor, the method comprising: Detecting a plurality of resonance points in a frequency characteristic of an input / output gain of the servo control device, the frequency characteristic being measured based on an input signal and an output signal having a changing frequency; setting a second plurality of filters in excess of the number of the first plurality of filters to suppress the plurality of resonance points; The filter adjustment method includes grouping the second plurality of filters to set the first plurality of filters. Effect of the Invention
[0008] According to each aspect of the present disclosure, it is possible to determine the allocation of filters so as to satisfy the constraint on the number of filters while achieving resonance suppression. [Brief description of the drawings]
[0009] [Figure 1] FIG. 2 is a block diagram showing a control system according to a first embodiment of the present disclosure. [Diagram 2] FIG. 1 is a block diagram showing an example of a filter formed by directly connecting a plurality of filters. [Diagram 3] 11A and 11B are diagrams illustrating an example of gain characteristics of input / output gain, filter settings for a plurality of resonance points, and grouping. [Figure 4] 3 is a block diagram showing a configuration of a control support unit including a frequency characteristic estimation unit in the first embodiment. FIG. [Diagram 5] FIG. 13 is a block diagram showing a model for calculating a reference model of input / output gain. [Figure 6] 10 is a characteristic diagram showing the frequency characteristics of the input / output gain of a servo control unit of a reference model and an estimated value of the frequency characteristics of the input / output gain of the servo control unit; FIG. [Figure 7] 11A and 11B are diagrams illustrating an example of gain characteristics of input / output gain, filter settings for a plurality of resonance points, and grouping. [Figure 8] FIG. 2 is a block diagram showing a configuration of a part of a control system focusing on a grouping unit. [Figure 9] FIG. 2 is a block diagram showing a machine learning unit of one embodiment of the present invention. [Figure 10] 2 is a flowchart showing an operation of a control support unit shown in FIG. 1. [Figure 11] FIG. 4 is a block diagram showing a control system of a second embodiment of the present disclosure. [Figure 12] FIG. 13 is a block diagram showing a modified example of the control system. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0010] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings.
[0011] (First embodiment) FIG. 1 is a block diagram showing a control system according to a first embodiment of the present disclosure. The control system 10 includes a servo control unit 100, a frequency generation unit 200, a frequency characteristic measurement unit 300, and a control assistance unit 400. The servo control unit 100 corresponds to a servo control device that controls a motor, the frequency characteristic measurement unit 300 corresponds to a frequency characteristic measurement device, and the control assistance unit 400 corresponds to a control assistance device. One or more of the frequency generating section 200, the frequency characteristic measuring section 300, and the control assisting section 400 may be provided within the servo control section 100. The frequency characteristic measuring section 300 may be provided within the control assisting section 400.
[0012] 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, the speed control unit 120, the filter 130, the current control unit 140, and the motor 150 configure a servo system of a speed feedback loop. The control support unit 400 detects multiple resonance points of the frequency characteristic of the input / output gain (amplitude ratio) of the servo control unit 100 obtained without the filter 130, sets multiple filters exceeding the constraint on the number of filters so as to suppress resonance at the multiple resonance points, and groups the multiple filters so that the number of filters is equal to or less than the constraint on the number of filters. The filter 130 is provided based on a combination of the grouped filters. Fig. 1 shows the servo control unit 100 after the filter 130 set by grouping is provided.
[0013] The motor 150 is a linear motor that moves in a straight line or a motor that has a rotating shaft. The object driven by the motor 150 is, for example, a mechanical part of a machine tool, a robot, or an industrial machine. The motor 150 may be provided as a part of the machine tool, the robot, the industrial machine, or the like. The control system 10 may be provided as a part of the machine tool, the robot, the industrial machine, or the like.
[0014] The subtractor 110 obtains the difference between the input speed command and the detected speed that is fed back, and outputs the difference to the speed control unit 120 as the speed deviation.
[0015] The speed control unit 120 performs PI control (Proportional-Integral Control), adds a value obtained by multiplying the speed deviation by integral gain K1v and integrating the result, and a value obtained by multiplying the speed deviation by proportional gain K2v, and outputs the result as a torque command to the filter 130. The speed control unit 120 includes a feedback gain. Note that the speed control unit 120 is not particularly limited to PI control, and other control, for example, PID control (Proportional-Integral-Differential Control) may be used. Equation 1 (hereinafter referred to as Equation 1) expresses the transfer function G V (s) is indicated.
number
[0016] The filter 130 is configured by connecting a plurality of filters that attenuate specific frequency components in series. Each filter is, for example, a notch filter, a low-pass filter, or a band-stop filter. In a machine such as a machine tool having a mechanism driven by a motor 150, there may be a plurality of resonance points, and each resonance may increase in the servo control unit 100. By connecting filters such as notch filters in series, each resonance of the plurality of resonance points can be reduced. The output of the filter 130 is output to the current control unit 140 as a torque command.
[0017] FIG. 2 is a block diagram showing an example in which the filter 130 is configured by directly connecting a plurality of filters. In FIG. 2, when there are k (k is a natural number of 2 or more) resonance points, the filter 130 is configured by connecting m (m is a natural number of 2 or more and m < k) filters 130-1 to 130-m in series. The m filters 130-1 to 130-m each correspond to a different frequency band. Hereinafter, the filter 130 will be described on the assumption that it is composed of m filters 130-1 to 130-m. The number m of filters constituting the filter 130 is limited because the number of filters that can be set in the servo control unit 100 is finite and there is a restriction on the number of filters. Therefore, it is difficult to provide the same number as the k resonance points. If the restriction on the number of filters is the maximum number of filters Nmax, the control support unit 400 determines the distribution of the m filters for the k resonance points so that the number m of filters is equal to or less than the maximum number of filters Nmax. Equation 2 (hereinafter referred to as Equation 2) shows the transfer function G F (s) of a notch filter as one of the filters 130, for example, the filter 130-1. The filters 130-2 to 130-m can also be each composed of a notch filter having a similar transfer function. Here, the coefficient δ in Equation 2 is the attenuation coefficient, the coefficient ω c is the central angular frequency, and the coefficient τ is the relative bandwidth. Assuming the center frequency is fc and the bandwidth is fw, the coefficient ω c is ω c = 2πfc, and the coefficient τ is represented by τ = fw / fc.
Equation
[0018] 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 . When motor 150 is a linear motor, the position of the movable part is detected by a linear scale (not shown) provided on motor 150, and a speed detection value is obtained by differentiating the position detection value. The obtained speed detection value is input to subtractor 110 as speed feedback. When the motor 150 is a motor having a rotating shaft, the rotational angle position is detected by a rotary encoder (not shown) provided on the motor 150, and the detected speed value is input to the subtractor 110 as speed feedback.
[0019] The servo control unit 100 is configured as described above. In addition to the servo control unit 100, the control system 10 further includes a frequency generating unit 200, a frequency characteristic measuring unit 300, and a control assisting unit 400. The frequency characteristic measuring unit 300 may be included in the control assisting unit 400.
[0020] The frequency generating section 200 outputs a sine wave signal as a speed command to the subtractor 110 and the frequency characteristic measuring section 300 of the servo control section 100 while changing the frequency.
[0021] The frequency characteristic measuring unit 300 measures the amplitude ratio (input / output gain) and phase delay between the input signal and the output signal for each frequency defined by the speed command, using the speed command (sine wave) that serves as an input signal generated by the frequency generating unit 200 and the detected speed (sine wave) that serves as an output signal output from a rotary encoder (not shown). Alternatively, the frequency characteristic measuring unit 300 measures the amplitude ratio (input / output gain) and phase delay between the input signal and the output signal for each frequency defined by the speed command, using the speed command (sine wave) that serves as an input signal generated by the frequency generating unit 200 and the differential of the detected position (sine wave) that serves as an output signal output from the linear scale.
[0022] When the servo control unit 100 operates in a state where the filter 130 is not provided, the frequency characteristic measuring unit 300 measures at least the frequency characteristics of the input / output gain and the phase delay, and outputs the frequency characteristics of the input / output gain and the phase delay to a resonance detection unit 401 and a frequency characteristic estimating unit 403 (described later) of the control support unit 400. When the servo control unit 100 operates in a state where some or all of the m filters 130-1 to 130-m of the filter 130 are provided, the frequency characteristic measuring unit 300 measures the frequency characteristics of the input / output gain and the phase delay, and outputs the frequency characteristics of the input / output gain and the phase delay to a grouping unit 404 (described later) of the control support unit 400.
[0023] 3, the servo control unit 100 operates in a state where the filter 130 is not provided, and the control support unit 400 detects k resonance points of the input / output gain output from the frequency characteristic measurement unit 300. Then, the control support unit 400 selects, for example, n, which exceeds the number of filters Nmax, which is a restriction on the number of filters, so as to suppress the resonance of the detected k resonance points. ini Each coefficient of the filters (coefficient ω in Equation 2) c , τ, δ) and make initial adjustments. ini Grouping of filters is performed to determine the number of filters to be initially adjusted, n ini The number of filters m(n ini The m filters are a first plurality of filters, n ini This filter becomes a second plurality of filters.
[0024] Number n ini may be less than, greater than, or equal to the number k. In other words, for k resonance points, multiple filters may be provided for one resonance point, or one filter may be provided for multiple resonance points. In other words, for k resonance points, k filters (n ini Even if the number of resonance points k is equal to or less than the number of filters Nmax, n iniIn some cases, it may be necessary to perform initial adjustment of each coefficient of n filters. If multiple filters are provided for one resonance point, ini >k, and the number of filters is n ini may exceed the number of filters Nmax. The number of filters n provided to suppress the resonance of k resonance points ini is set to a number that exceeds the number of filters Nmax, which is the constraint on the number of filters, but ini Even if is less than or equal to the number of filters Nmax, n ini The number of filters m(n ini >m) may be set, because there are cases where a smaller number of filters is better even if the objective function such as the cutoff frequency deteriorates.
[0025] Fig. 3 is a diagram showing an example of gain characteristics of input / output gain, filter settings for a plurality of resonance points, and grouping. In Fig. 3, eight resonance points P1 to P8 are generated, the coefficients of the eight filters are initially adjusted so as to suppress the resonance points P1 to P8, and one filter is set for each of the resonance points P2 and P3, the resonance points P5 and P6, and the resonance points P7 and P8 by grouping.
[0026] The configuration and operation of the control support unit 400 will be described in detail below. <Control Support Unit 400> As shown in FIG. 1, the control support unit 400 includes a resonance detection unit 401 , a filter setting unit 402 , a frequency characteristic estimation unit 403 , and a grouping unit 404 .
[0027] When the servo control unit 100 operates without the filter 130 being provided, the resonance detection unit 401 acquires 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. The resonance detection unit 401 may detect an anti-resonance point in addition to the resonance point. By detecting the anti-resonance point, the filter setting unit 402 determines n ini pieces(n ini When setting the range of attenuation center frequencies of each filter of the filters (>m), the range of attenuation center frequencies can be set between the frequencies of the anti-resonance points.
[0028] The filter setting unit 402 sets n ini The filters are set and cooperate with the frequency characteristic estimation unit 403 to estimate n ini The coefficients of the filters are initially adjusted for each filter. The grouping unit 404 uses the initially adjusted n ini The filters are grouped to set m filters (k>m).
[0029] The initial filter adjustment by the filter setting section 402 and the grouping by the grouping section 404 will now be described. As mentioned above, the number of filters to be initially adjusted is n ini may be greater than or less than the number of filters Nmax, which is a restriction on the number of filters. ini may be greater than or equal to the number k of resonance points. The number of filters m is equal to or less than the number Nmax of filters, and the number of filters n ini Less. In the following explanation, the number of filters to be initially adjusted is n ini exceeds the number of filters Nmax and is equal to the number of resonance points k (n ini >Nmax, n ini = k). Number of filters to be initially adjusted n ini exceeds the number of filters Nmax, so in the following description, the frequency characteristic estimating section 403 is used to estimate the frequency characteristics of the input / output gains. However, the number of filters to be initially adjusted, n inidoes not exceed the number of filters Nmax, the frequency characteristics of the input / output gains may be estimated using the frequency characteristic estimation unit 403, or the servo control unit 100 may actually be operated and the frequency characteristics of the input / output gains may be measured by the frequency characteristic measurement unit 300.
[0030] (Initial filter adjustment) Below, k filters (k=n ini ) coefficient ω c , τ, and δ for each filter, and the configuration and operation of frequency characteristic estimation section 403 will be described. The filter setting unit 402 selects one resonance point from the k resonance points one by one, and sets each coefficient ω of one filter to attenuate the resonance of the selected resonance point. c , τ, and δ are set and output to the frequency characteristic estimation unit 403 in sequence.
[0031] The frequency characteristic estimation unit 403 estimates each coefficient ω of one filter acquired from the filter setting unit 402. c , τ, δ, and the frequency characteristics of the input / output gain and phase delay obtained from the frequency characteristic measuring unit 300 when the servo control unit 100 operates without the filter 130 being provided, an estimate of the frequency characteristics of the input / output gain of the servo control unit 100 when one filter is set to the filter 130 is calculated, and output to the filter setting unit 402.
[0032] 4 is a block diagram showing the configuration of a control support unit including a frequency characteristic estimating unit 403. In FIG. 4, frequency characteristic measuring unit 300 is also shown for the purpose of explaining the operation of frequency characteristic estimating unit 403. As shown in FIG. 4, the frequency characteristic estimation unit 403 includes a filter information acquisition unit 4031 , a pre-adjustment state storage unit 4032 , a frequency characteristic calculation unit 4033 , and a state estimation unit 4034 .
[0033] The filter information acquisition unit 4031 acquires each coefficient ω of one filter from the filter setting unit 402. c , τ, and δ are obtained and output to the frequency characteristic calculation unit 4033.
[0034] As described above, the pre-adjustment state storage unit 4032 stores the frequency characteristic C1 of the input / output gain and phase delay acquired from the frequency characteristic measurement unit 300 when the servo control unit 100 operates in a state in which the filter 130 is not provided.
[0035] The frequency characteristic calculation unit 4033 receives each coefficient ω of one filter from the filter information acquisition unit 4031. c , τ, and δ, and reads out the frequency characteristic C 1 from the pre-adjustment state storage unit 4032 . Then, the frequency characteristic calculation unit 4033 calculates each coefficient ω c , τ, δ, the transfer function G F Using (jω), calculate the frequency characteristic C2 of the input / output gain and phase delay of one filter.
[0036] Then, the frequency characteristic calculation unit 4033 outputs to the state estimation unit 4034 the measured frequency characteristic C1 of the input / output gain and phase delay of the servo control unit 100 and the calculated frequency characteristic C2 of the input / output gain and phase delay of the filter.
[0037] The state estimation unit 4034 adds the frequency characteristics C1 and C2 to obtain an estimate E (E=C1+C2) of the frequency characteristics of the input / output gain and phase delay of the servo control unit 100 provided with one filter, and outputs the estimate E to the filter setting unit 402. Hereinafter, of the estimate values E of the frequency characteristics of the input / output gain and the phase delay, the estimate value of the frequency characteristics of the input / output gain is referred to as E1, and the estimate value of the frequency characteristics of the phase delay is referred to as E2. By providing one filter in the servo control unit 100, the filter setting unit 402 can determine whether the resonance at the selected resonance point has attenuated to an acceptable range based on an estimated value E1 of the frequency characteristic of the input / output gain of the servo control unit 100.
[0038] The filter setting unit 402 determines whether the resonance at the selected resonance point has attenuated to an allowable range as follows. The filter setting unit 402 stores in advance a reference model of the input / output gain of the servo control unit 100. The reference model is a model of a servo control unit having ideal characteristics without resonance. For example, the reference model is a model of the inertia Ja and torque constant K of the model shown in FIG. t , proportional gain K p , integral gain K I , differential gain K D The inertia Ja can be calculated by the following equation. The inertia Ja is the sum of the motor inertia and the mechanical inertia.
[0039] 6 is a characteristic diagram showing the frequency characteristics of the input / output gain of the servo control unit of the reference model and an estimated value of the frequency characteristics of the input / output gain of the servo control unit 100. As shown in the characteristic diagram of FIG. 6, the reference model has region A, which is a frequency domain where the ideal input / output gain is equal to or greater than a certain input / output gain, for example, equal to or greater than -20 dB, and region B, which is a frequency domain where the ideal input / output gain is less than the certain input / output gain. In region A of FIG. 6, the ideal input / output gain of the reference model is shown by curve MC1 (thick line). In region B of FIG. 6, the ideal virtual input / output gain of the reference model is shown by curve MC 11 The input / output gain of the reference model is constant and the line MC 12 6, the curves of the estimated value E1 of the input / output gain of the servo control unit are indicated by curves RC1 and RC2, respectively.
[0040] In region A, if the curve RC1 of the estimated value E1 of the input / output gain does not exceed the curve MC1 of the ideal input / output gain of the reference model in a band centered on the selected resonance point, the filter setting unit 402 determines that the resonance of the selected resonance point has attenuated to within the tolerable range, and if it does exceed the curve MC1, it determines that the resonance of the selected resonance point has not attenuated to within the tolerable range. In the region B exceeding the frequency where the input / output gain becomes sufficiently small, the curve RC1 of the estimated input / output gain E1 is the curve MC of the ideal virtual input / output gain of the reference model. 11 Therefore, in region B, as mentioned above, the input / output gain of the reference model is the ideal gain characteristic curve MC11 Instead, a linear MC with a constant input / output gain (for example, -20 dB) 12 The filter setting unit 402 uses a curve RC1 of the estimated value E1 of the input / output gain in a band centered on the selected resonance point, which is in agreement with a straight line MC of the constant-value input / output gain. 12 If it does not exceed the tolerance, it is determined that the resonance of the selected resonance point has attenuated to the tolerance range, and if it exceeds the tolerance, it is determined that the resonance of the selected resonance point has not attenuated to the tolerance range since there is a possibility of instability.
[0041] When the resonance of the selected resonance point is attenuated to within the allowable range, the filter setting unit 402 sets each coefficient ω c , τ, and δ. The filter setting unit 402 stores each coefficient ω c , τ, and δ are changed and output to the frequency characteristic estimation unit 403, and the above-mentioned process is repeated. In this way, the filter setting unit 402 determines each coefficient ω of one filter such that the resonance of the selected resonance point is attenuated to an allowable range. c , τ, δ are obtained.
[0042] The filter setting unit 402 selects each coefficient ω of one filter so that the resonance of the selected resonance point is attenuated to an allowable range. c The cutoff frequency is calculated from the estimated value E2 of the frequency characteristic of the phase delay corresponding to each set of , τ, and δ, and the coefficient ω c , τ, δ are the coefficients of one filter, ω c , τ, δ are determined. The cutoff frequency is, for example, a frequency at which the gain characteristic of a Bode diagram obtained by measuring a frequency response calculated from the input / output gain of the servo control unit 100 is −3 dB, or a frequency at which the phase characteristic is −180 degrees. As the cutoff frequency increases, the feedback gain increases and the response speed becomes faster. Next, the filter setting unit 402 selects another resonance point from the k resonance points, and calculates each coefficient ω c , τ, and δ are set and output to the frequency characteristic estimation unit 403.
[0043] The filter setting unit 402 and the frequency characteristic estimating unit 403 cooperate to calculate the coefficients ω c The operation of determining , τ, and δ is performed for k filters in sequence, and the coefficients ω c , τ, and δ are determined, and the coefficients of the k filters are initially adjusted for each filter. The filter setting unit 402 outputs the initially adjusted k filter coefficients to the grouping unit 404 .
[0044] In the above description, the servo control unit 100 is operated without the filter 130 to obtain the frequency characteristics of the input / output gain (amplitude ratio) and the phase delay to detect the resonance point. However, the frequency characteristics of the input / output gain and the phase delay in the absence of the filter 130 may be obtained by other methods. For example, the coefficients ω c , τ, and δ are used to calculate the frequency characteristics of the input / output gain and phase delay of the filter 130. Then, the servo control unit 100 provided with the filter 130 is operated to obtain the frequency characteristics of the input / output gain and phase delay, and the frequency characteristics of the input / output gain and phase delay of the filter 130 are subtracted from these frequency characteristics, thereby making it possible to obtain the frequency characteristics of the input / output gain and phase delay in the case where the filter 130 is not present.
[0045] (Filter grouping) Next, an operation of grouping section 404 grouping k initially adjusted filters to set m filters (k>m, Nmax≧m) not greater than Nmax, which is a constraint on the number of filters, will be described. In order to group the initially adjusted k filters into m (k>m) filters, the grouping unit 404 sets a filter combination in which, among the k filters, there is one or more filters that combine adjacent filters into one. Alternatively, the grouping unit 404 sets a filter combination in which, among the k filters, there is one or more sets of two or more filters that combine adjacent three or more filters into two or more that is less than the number of adjacent three or more filters. As a method for grouping k filters into m filters, for example, there are the following four methods. In the following explanation, an example will be described in which one or more filters (hereinafter, this filter will be referred to as a composite filter) are provided that combines multiple adjacent filters into one in order to convert the k initially adjusted filters into m (k>m) filters.
[0046] (1) Exhaustive search of filter combinations By combining multiple adjacent filters into one composite filter, the number of filter combinations that combine k filters into m filters is the number of (m-1) partitions selected from the (k-1) partitions that separate the k filters, so (k-1) C (m-1) It becomes one. The grouping unit 404 (k-1) C (m-1) The combination of filters is calculated, and the coefficient ω of the composite filter is calculated for each combination of filters. c , τ, and δ. For one filter combination, the grouping unit 404 sets the filter combination including one composite filter as the filter 130 in the servo control unit 100. Then, similar to the operation of the filter setting unit 402, the grouping unit 404 adjusts the coefficients of one composite filter so that the frequency characteristics of the input / output gain obtained by operating the servo control unit 100 do not exceed the curve of the ideal input / output gain frequency characteristics of the reference model described with reference to Figs. 5 and 6 and the cutoff frequency becomes the largest.
[0047] When there are other composite filters in addition to the one composite filter in one filter combination, the grouping unit 404 adjusts the coefficients of the other composite filters in the same manner. A filter combination in which the coefficients of one or more composite filters are adjusted in this manner becomes one grouping candidate. When the grouping unit 404 finds one grouping candidate, it finds grouping candidates for other combinations.
[0048] The grouping unit 404 repeats the above operations to (k-1) C (m-1) We searched for individual grouping candidates and (k-1) C (m-1) The value of an objective function such as a cutoff frequency for each of the groupings is calculated, and the grouping candidate that provides the best objective function value is selected as the optimal combination. If the objective function is a cutoff frequency, the grouping candidate that provides the highest cutoff frequency is selected as the optimal combination.
[0049] Fig. 7 is a diagram showing an example of the gain characteristics of input / output gain, filter settings for a plurality of resonance points, and grouping. In Fig. 7, an example is shown in which 10 resonance points are generated, the coefficients of 10 filters are initially adjusted to suppress the 10 resonance points, and the 10 filters are grouped into 5 filters. In Fig. 7, group G1 and group G2 are examples of grouping of different combinations of filters. When performing an exhaustive search of filter combinations, the number of filter combinations in which 10 filters are grouped into 5 filters, as shown in Fig. 7, is 9C4. Also, the number of filter combinations in which 10 filters are grouped into 5 or fewer filters is (9C4+9C3+9C2+9C1).
[0050] (2) Machine learning exploration When performing a search by machine learning, the grouping unit 404 is made up of a grouping determination unit 410 and a machine learning unit 420, as shown in Fig. 8. Fig. 8 is a block diagram showing the configuration of a part of the control system centered around the grouping unit. The machine learning unit 420 sets a filter combination including one or more composite filters that combine multiple adjacent filters into one among the k filters, in order to set the k filters initially adjusted by the filter setting unit 402 into m filters (k>m). Note that the filter combination may be set by the grouping determination unit 410, which may then output the filter combination to the machine learning unit 420.
[0051] The machine learning unit 420 first acquires coefficients of k filters that have been initially adjusted from the filter setting unit 402 via the grouping determination unit 410. The machine learning unit 420 then provides one or more composite filters, and in one combination of k filters that have been initially adjusted as m filters, sets a filter combination including one composite filter as filter 130 in the servo control unit 100. The machine learning unit 420 then acquires 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 performs machine learning in the same manner as the operation of the filter setting unit 402 so that the frequency characteristics of the acquired input / output gain do not exceed the curve of the ideal input / output gain frequency characteristics of the reference model described with reference to FIG. 5 and FIG. 6 and the cutoff frequency becomes the largest, and calculates the coefficient ω of one composite filter. c , τ, and δ are adjusted. The operation of setting a filter combination including one composite filter in filter 130 and the operation of acquiring the frequency characteristics of input / output gain and phase delay from frequency characteristic measurement unit 300 are performed via grouping determination unit 410. Note that machine learning unit 420 may directly set a filter combination including one composite filter in filter 130 and acquire the frequency characteristics of input / output gain and phase delay from frequency characteristic measurement unit 300 without going through grouping determination unit 410. In a filter combination, if there are other composite filters in addition to the one composite filter described above, the coefficients of the other composite filters are also adjusted in the same manner. A filter combination in which the coefficients of one or more composite filters have been adjusted in this manner becomes one grouping candidate.
[0052] An example of the above operation of the machine learning unit 420 will be specifically described with reference to Fig. 3. Here, Fig. 3 is used as an explanatory diagram for determining one grouping candidate. In the gain characteristics of the input / output gain, eight resonance points P1 to P8 occur, and the machine learning unit 420 sets filter combinations in which one composite filter is set for each of the resonance points P2 and P3, the resonance points P5 and P6, and the resonance points P7 and P8, for example. The machine learning unit 420 sets, for example, a filter combination including one composite filter set for the resonance points P2 and P3 to the filter 130. At this time, no composite filter is set for the resonance points P5 and P6 and the resonance points P7 and P8. The machine learning unit 420 acquires the frequency characteristic of the input / output gain obtained by operating the servo control unit 100 including the filter 130 equipped with the composite filter set for the resonance points P2 and P3. The machine learning unit 420 performs machine learning so that the frequency characteristic of the acquired input / output gain does not exceed the curve of the ideal input / output gain frequency characteristic of the reference model described with reference to Figs. 5 and 6 and the cutoff frequency becomes the largest, in the same manner as the operation of the filter setting unit 402, and adjusts the coefficient of the composite filter set for the resonance points P2 and P3. Thereafter, the machine learning unit 420 similarly adjusts the coefficients of the composite filter set for the resonance points P5 and P6 and the coefficients of the composite filter set for the resonance points P7 and P8. In this way, a combination of filters in which the coefficients of the three composite filters provided for the resonance points P2 and P3, the resonance points P5 and P6, and the resonance points P7 and P8 have been adjusted is obtained as one grouping candidate.
[0053] When the machine learning unit 420 obtains one grouping candidate, it provides one or more composite filters and obtains another grouping candidate in which the initially adjusted k filters are replaced by m filters. By repeating this operation, the machine learning unit 420 obtains each coefficient ω c , τ, and δ are sequentially output to the grouping determination unit 410. The configuration and operation of the machine learning unit 420 will be described in detail later.
[0054] The grouping determination unit 410 sets the filter combination of one grouping candidate obtained by the machine learning unit 420 to the filter 130, operates the servo control unit 100, and determines an objective function such as a cutoff frequency for one grouping candidate using the frequency characteristic of the phase delay of the servo control unit 100.
[0055] The grouping determination unit 410 determines an objective function such as a cutoff frequency for the next grouping candidate obtained by the machine learning unit 420 in a manner similar to the operation for determining the cutoff frequency for one grouping candidate. The grouping determination unit 410 repeats this operation to determine, from among the multiple grouping candidates determined by the machine learning unit 420, a combination of filters that provides the best value for an objective function such as a cutoff frequency.
[0056] (3) Search based on set rules The grouping unit 404 groups the initially adjusted k filters into m (k>m) filters based on a determined rule. The rule may be, for example, not grouping filters with small damping (damping ratio) together (= grouping weak filters together), grouping filters starting from high frequency filters, not grouping filters with far-apart damping center frequencies, etc. A method for obtaining a composite filter by combining a plurality of filters into one is performed in the same manner as the operation for obtaining grouping candidates described in the exhaustive search of the combination of filters in (1) above. When a plurality of grouping candidates are obtained based on this rule, the grouping unit 404 obtains, from among the obtained grouping candidates, a combination of filters that provides the best value for an objective function such as a cutoff frequency.
[0057] (4) Group filters together The grouping unit 404 groups the initially adjusted k filters one by one to obtain m (k>m) filters. For example, the k filters are reduced by one to obtain (k-1) filters, the (k-1) filters are reduced by one to obtain (k-2) filters, and so on until there are m filters, so that (k-1) C (m-1) This eliminates the need to evaluate the pattern.
[0058] A method for obtaining a composite filter by combining a plurality of filters into one is as follows. The grouping unit 404 operates the servo control unit 100 in which (k-1) filters, including one composite filter formed by grouping adjacent filters, are set as the filter 130. In the same manner as the operation of the filter setting unit 402, the grouping unit 404 sets the coefficient ω of one composite filter so that the frequency characteristic of the input / output gain obtained by operating the servo control unit 100 does not exceed the curve of the ideal input / output gain frequency characteristic of the reference model described with reference to FIGS. 5 and 6 and the cutoff frequency becomes the largest. c , τ, and δ are adjusted. Next, the grouping unit 404 newly provides a composite filter that groups adjacent filters among the (k-1) filters to provide (k-2) filters, and adjusts the coefficients of the newly provided composite filter so that the frequency characteristics of the input / output gain obtained by operating the servo control unit 100 in which the (k-2) filters are set as filters 130 do not exceed the curve of the ideal input / output gain frequency characteristics of the reference model described using Figures 5 and 6, as in the operation of the filter setting unit 402, and the cutoff frequency is maximized. These operations are repeated to obtain m combinations of filters.
[0059] <Machine Learning Department> The configuration and operation of the machine learning unit 420 when the grouping unit 404 determines grouping candidates for the filter 130 by the machine learning described above in (2) will be described below. In the following description, the machine learning unit 420 performs reinforcement learning; however, the machine learning is not limited to reinforcement learning, and for example, the machine learning unit 420 may perform supervised learning.
[0060] In the filter setting unit 402, the machine learning unit 420 sets a combination of filters including one or more composite filters that combine multiple adjacent filters into one, in order to turn the initially adjusted k filters into m (k>m) filters. The machine learning unit 420 provides one or more composite filters, and in one combination of k initially adjusted filters as m filters, sets a filter combination including one composite filter as filter 130 in the servo control unit 100. The machine learning unit 420 operates the servo control unit 100 to obtain, as a state S, frequency characteristics of input / output gain and phase delay obtained from the frequency characteristic measuring unit 300. Then, the machine learning unit 420 calculates each coefficient ω of one composite filter of filter 130 of the servo control unit 100 related to the state S. c Q-learning is performed by adjusting the values of , τ, and δ as action A. If there are other composite filters in addition to the one composite filter described above in one filter combination, Q-learning is also performed on the coefficients of the other composite filters in the same manner. As is well known to those skilled in the art, the purpose of Q-learning is to select, as the optimal action, the action A with the highest value Q(S,A) from among the actions A that can be taken in a certain state S.
[0061] Specifically, an agent (machine learning device) selects various actions A in a certain state S, and learns the correct value Q(S, A) by choosing the better action based on the reward given for that action A.
[0062] Also, we want to maximize the total rewards we can get in the future, so we ultimately want Q(S,A)=E[Σ(γ t )r t Here, E[] represents the expected value, t is time, γ is a parameter called the discount rate, which will be described later, and r t is the reward at time t, and Σ is the sum at time t. The expected value in this equation is the expected value when the state changes according to the optimal action. The update equation for such value Q(S,A) can be expressed, for example, by the following equation 3 (hereinafter referred to as equation 3).
[0063]
number
[0064] In the above formula 3, S t represents the state of the environment at time t, and A t represents the action at time t. Action A t Therefore, the condition is S t+1 Changes to r t+1 represents the reward obtained by changing the state. Also, the term with max represents the reward obtained by changing the state S t+1 It is calculated by multiplying the Q value of the action A with the highest Q value known at that time by γ. Here, γ is a parameter with a range of 0<γ≦1 and is called the discount rate. Also, α is a learning coefficient with a range of 0<α≦1.
[0065] The above formula 3 is Trial A t As a result, the reward r t+1 Based on this, state S t Action A in t The value of Q(S t ,A t ) to update the
[0066] The machine learning unit 420 observes state information S including frequency characteristics of input / output gain and phase delay for each frequency output from the frequency characteristic measuring unit 300, and determines an action A. The machine learning unit 420 receives a reward every time it performs an action A. The reward will be described later. In Q-learning, the machine learning unit 420 searches for an optimal action A that maximizes the total reward in the future by trial and error, for example. In this way, the machine learning unit 420 selects an optimal action A (optimal coefficients ω c , τ, δ) can be selected.
[0067] FIG. 9 is a block diagram illustrating the machine learning unit 420 of one embodiment of the present invention. In order to perform the above-mentioned reinforcement learning, as shown in FIG. 9, 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.
[0068] The state information acquisition unit 421 receives the initially adjusted coefficients ω c , τ, and δ are acquired from the filter setting unit 402. In addition, in order to convert the initially adjusted k filters into m (k>m) filters, the state information acquiring unit 421 sets a filter combination having one or more composite filters that combine adjacent filters in the k filters. The coefficient ω of the composite filter at the time when Q-learning is first started is c The initial values of , τ, and δ are generated in advance by the user. The coefficients of the filters other than the coefficients of the composite filter are initially adjusted values. In this embodiment, the coefficients ω of the composite filter generated by the user are c The initial settings of,τ,and,δ,are adjusted to optimal ones by reinforcement learning.
[0069] When Q-learning is first started, the state information acquisition unit 421 outputs to the behavior information generation unit 4223 information specifying adjacent filters to be removed from among the k initially adjusted filters, and the initial values of the coefficients of the composite filter to be replaced.
[0070] The state information acquisition unit 421 also acquires the coefficients ω c , τ, and δ, the state S including the input / output gain (amplitude ratio) and phase delay obtained by driving the servo control unit 100 using a speed command (sine wave) is acquired from the frequency characteristic measuring unit 300 via the grouping determining unit 410, and output to the learning unit 422. This state information S corresponds to the environmental state S in Q-learning.
[0071] The learning unit 422 is a part that learns a value Q(S, A) when a certain action A is selected under a certain environmental state S. The learning unit 422 includes a reward output unit 4221, a value function update unit 4222, and a behavior information generation unit 4223.
[0072] The reward output unit 4221 is a part that calculates a reward when an action A is selected in a certain state S. First, the servo control unit 100 operates having the filter 130 whose composite filter coefficient is an initial value. The reward output unit 4221, like the operation of the filter setting unit 402, gives a negative reward when the frequency characteristic of the input / output gain among the frequency characteristics of the input / output gain and the phase delay output from the frequency characteristic measuring unit 300 exceeds the curve of the ideal input / output gain frequency characteristic of the reference model described using Fig. 5 and Fig. 6, and corrects the composite filter coefficient from the initial value. On the other hand, when the input / output gain is equal to or less than the input / output gain value of the reference model, the reward output unit 4221 obtains a first cutoff frequency that is the value of the first objective function by using the frequency characteristic of the phase delay output from the frequency characteristic measurement unit 300. The cutoff frequency is, for example, a frequency at which the gain characteristic of the Bode diagram obtained by measuring the frequency response calculated from the input / output gain of the servo control unit 100 is −3 dB, or a frequency at which the phase characteristic is −180 degrees. As the cutoff frequency increases, the feedback gain increases and the response speed becomes faster.
[0073] Next, the servo control unit 100 having the filter 130 with the adjusted initial value of the composite filter coefficient operates. The reward output unit 4221, like the operation of the filter setting unit 402, gives a negative reward when the frequency characteristic of the input / output gain among the frequency characteristics of the input / output gain and the phase delay output from the frequency characteristic measuring unit 300 exceeds the curve of the ideal input / output gain frequency characteristic of the reference model described using Fig. 5 and Fig. 6, and further corrects the composite filter coefficient. On the other hand, when the input / output gain is equal to or less than the value of the input / output gain of the reference model, the reward output unit 4221 uses the frequency characteristic of the phase delay output from the frequency characteristic measurement unit 300 to find a second cutoff frequency that is the value of the second objective function.
[0074] The reward output unit 4221 obtains the difference between the first cutoff frequency and the second cutoff frequency, which is the evaluation function. The reward output unit 4221 then gives a positive reward if the second cutoff frequency is greater than the first cutoff frequency, a zero reward if the second cutoff frequency is the same as the first cutoff frequency, and a negative reward if the second cutoff frequency is smaller than the first cutoff frequency.
[0075] Next, the reward output unit 4221 operates in a similar manner to that for determining the second cutoff frequency, and the servo control unit 100 having a filter 130 whose coefficients of the composite filter have been further adjusted operates to determine a third cutoff frequency, which is the value of the third objective function, using the frequency characteristic of the phase delay output from the frequency characteristic measuring unit 300 when the input / output gain is less than or equal to the value of the input / output gain of the reference model. The reward output unit 4221 obtains the difference between the second cutoff frequency and the third cutoff frequency, which is the evaluation function. The reward output unit 4221 then gives a positive reward if the third cutoff frequency is greater than the second cutoff frequency, a zero reward if the third cutoff frequency is the same as the second cutoff frequency, and a negative reward if the third cutoff frequency is smaller than the second cutoff frequency.
[0076] Furthermore, the reward output unit 4221 repeatedly performs an operation similar to the above-described operation of giving a reward using the second cutoff frequency and the third cutoff frequency when the input / output gain is equal to or less than the value of the input / output gain of the reference model, by adjusting the coefficients of the composite filter. The reward output unit 4221 has been described above.
[0077] Value function update unit 4222 updates value function Q stored in value function memory unit 424 by performing Q-learning based on state S, action A, state S' when action A is applied to state S, and the reward calculated as described above. The value function Q may be updated by online learning, batch learning, or mini-batch learning. Online learning is a learning method in which a certain action A is applied to the current state S, and the value function Q is updated immediately each time the state S transitions to a new state S'. Meanwhile, batch learning is a learning method in which a certain action A is applied to the current state S, and the state S transitions to a new state S' repeatedly, thereby collecting learning data, and updating the value function Q using all of the collected learning data. Furthermore, mini-batch learning is a learning method intermediate between online learning and batch learning, in which the value function Q is updated each time a certain amount of learning data is accumulated.
[0078] The behavior information generation unit 4223 selects behavior A in the Q-learning process for the current state S. In the Q-learning process, the behavior information generation unit 4223 selects each coefficient ω c , τ (corresponding to action A in Q-learning), the action information A is generated and the generated action information A is output to the action information output unit 423.
[0079] In addition, the behavior information generation unit 4223 may take a measure to select behavior A' by a known method such as a greedy method that selects behavior A' with the highest value Q(S, A) among the currently estimated values of behavior A, or an ε-greedy method that randomly selects behavior A' with a certain small probability ε and otherwise selects behavior A' with the highest value Q(S, A).
[0080] The behavior information output unit 423 is a part that transmits 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 current state S, that is, the currently set coefficients ω c , τ, and δ are adjusted to transition to the next state S′.
[0081] Value function storage unit 424 is a storage device that stores value function Q. Value function Q may be stored as a table (hereinafter referred to as an action value table) for each state S and action A, for example. Value function Q stored in value function storage unit 424 is updated by value function update unit 4222. Furthermore, value function Q stored in value function storage unit 424 may be shared with other machine learning units 420. If value function Q is shared by multiple machine learning units 420, reinforcement learning can be performed in a distributed manner in each machine learning unit 420, thereby improving the efficiency of reinforcement learning.
[0082] The above explanation is about reinforcement learning when one composite filter is provided in filter 130, but if filter 130 has other composite filters in addition to the one composite filter, reinforcement learning is similarly performed for the coefficients of the other composite filters. Then, one or more composite filters are provided, and reinforcement learning is similarly performed when there are other combinations in which the initially adjusted k filters are used as m filters.
[0083] The grouping candidate output unit 425 calculates, based on the value function Q updated by the value function update unit 4222 through Q-learning, each coefficient ω of each filter of a filter combination including one or more composite filters that maximizes the value Q(S,A). c , τ, and δ are calculated, and the coefficients of this filter combination are output to the grouping determination unit 410 as grouping candidates. More specifically, the grouping candidate output unit 425 acquires the value function Q stored in the value function storage unit 424. This value function Q is updated by the value function update unit 4222 performing Q-learning as described above. Then, based on the value function Q, the grouping candidate output unit 425 calculates the coefficients ω of each filter of a combination of filters including one or more composite filters that maximizes the value Q(S,A). c , τ, and δ are calculated, and the coefficients of this filter combination are output to the grouping determination unit 410 as grouping candidates. One or more composite filters are provided, and reinforcement learning is performed for other combinations of the initially adjusted k filters into m filters, and each coefficient ω of each filter in the other combination of filters including one or more composite filters that maximizes the value Q(S,A) is calculated. c , τ, and δ are calculated. The grouping candidate output unit 425 outputs the coefficients of this filter combination as a next grouping candidate to the grouping determination unit 410. For example, if the group G1 shown in Fig. 7 is one filter combination, the group G2 is an example of another filter combination.
[0084] As described above, the machine learning unit 420 calculates each coefficient ω c ,The optimal values of,τ,and,δ,are learned.
[0085] The functional blocks included in the control system 10 have been described above. To realize these functional blocks, the control system 10, the servo control unit 100, or the control assistance unit 400 includes a processor such as a CPU (Central Processing Unit). The control system 10, the servo control unit 100, or the control assistance unit 400 also includes a main memory such as an auxiliary storage device such as an HDD (Hard Disk Drive) that stores various control programs such as application software or an OS (Operating System), and a RAM (Random Access Memory) for storing data temporarily required for the processor to execute a program.
[0086] In the control system 10, the servo control unit 100, or the control assistance unit 400, the arithmetic processing unit reads the application software or OS from the auxiliary storage device, and performs arithmetic processing based on the application software or OS while expanding the read application software or OS in the main storage device. The arithmetic processing unit also controls various hardware devices provided in each device based on the results of this calculation. In this way, the functional blocks of this embodiment are realized. In other words, this embodiment can be realized by the cooperation of hardware and software.
[0087] When the amount of calculations required for the control assistance unit 400 is large, for example, a personal computer can be equipped with a GPU (Graphics Processing Units) to use the GPU for calculations and perform high-speed processing using a technology called GPGPU (General-Purpose computing on Graphics Processing Units). Furthermore, in order to perform processing at a higher speed, a computer cluster can be constructed using multiple computers equipped with such GPUs, and parallel processing can be performed by multiple computers included in the computer cluster.
[0088] Next, the operation of the control support unit 400 will be described with reference to a flowchart. Fig. 10 is a flowchart showing the operation of the control support unit.
[0089] In step S11, the resonance detection unit 401 acquires, from the frequency characteristic measurement unit 300, the frequency characteristic of the input / output gain (amplitude ratio) of the servo control unit 100 when the servo control unit 100 operates without the filter 130 being provided. In step S12, the resonance detection unit 401 detects k resonance points of the frequency characteristics (gain characteristics) of the input / output gains.
[0090] In step S13, the filter setting unit 402 cooperates with the frequency characteristic estimating unit 403 to set k filters corresponding to the k resonance points, and initially adjusts the coefficients of the k filters for each filter. Number of filters to be initially adjusted n ini is set to the same number as the number of resonance points (k=n ini ), n is a number smaller than the number of resonance points k ini (k>n ini ) filters may be initially adjusted. In step S14, the grouping unit 404 groups the initially adjusted k filters to set m filters (k>m).
[0091] In step S15, the control support unit 400 determines whether to continue the process of grouping k filters and setting m filters (k>m), and if so, returns to step S14, and if not, terminates the operation of the control support unit.
[0092] According to the embodiment described above, it is possible to determine the allocation of filters so as to satisfy the constraint on the number of filters while realizing resonance suppression.
[0093] Second Embodiment In the first embodiment, when measuring the frequency characteristics of the input / output gain (amplitude ratio) and phase delay of the servo control unit 100, the frequency characteristic measuring unit 300 calculates the frequency characteristics from the speed command and speed feedback, which are sine wave signals whose frequencies change. In this embodiment, the frequency generating unit 200 inputs a sine wave signal while changing the frequency to the upstream of the current control unit 140. Then, when measuring the frequency characteristics of the input / output gain and phase delay of the servo control unit 100, the frequency characteristic measuring unit 300 calculates the frequency characteristics from the sine wave signal input to the upstream of the current control unit 140 and the output of the speed control unit 120.
[0094] Fig. 11 is a block diagram showing a control system according to the second embodiment of the present disclosure, in which the same components as those of the control system 10 shown in Fig. 1 are denoted by the same reference numerals and will not be described. As shown in Fig. 11, in the control system 10A, an adder 160 is provided in front of a subtractor 170, and a sine wave signal whose frequency changes and is output from a frequency generating 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 an amplifier 180. The amplifier 180 includes a current detector, and a current detected by the current detector is input to the subtractor 170. The subtractor 170, the current control unit 140, and the amplifier 180 form a current feedback loop, and the current feedback loop is included in the speed feedback loop. The sine wave signal corresponds to a first signal whose frequency changes, and the output of the filter 130 corresponds to a second signal input to the current feedback loop in the speed feedback loop.
[0095] The inductance of motor 150 changes nonlinearly depending on the current flowing through motor 150 due to the influence of magnetic saturation and the like. When the servo parameters are changed from pre-adjustment to post-adjustment servo parameters, the torque command input to current control unit 140 changes, and if the current gain of current control unit 140 is constant, the current flowing through motor 150 also changes. When the current flowing through motor 150 changes and the inductance changes nonlinearly, the characteristics of the current feedback loop also change nonlinearly.
[0096] In this embodiment, the level of the input signal input to the subtractor 110 is set to zero, the frequency generating unit 200 inputs a sine wave signal while changing the frequency to the front stage of the current control unit 140, and the frequency characteristic measuring unit 300 measures the frequency characteristics of the input / output gain and phase delay of the servo control unit 100 from this sine wave signal and the output of the speed control unit 120. In this way, the input to the current feedback loop becomes constant, so that the control support unit 400 can find multiple resonances while maintaining the linearity of the current feedback loop characteristics.
[0097] The first and second embodiments have been described above. In the first and second embodiments, the filter setting unit 402 may include a machine learning unit having a configuration similar to that of the machine learning unit 420 shown in FIG. 9, and the coefficients of the k filters may be initially adjusted for each filter using the machine learning unit. The operation of initially adjusting each filter may include: c , τ, and δ are similar to those for adjusting the phase difference.
[0098] In each of the above-described embodiments, the objective function is not limited to the cutoff frequency, but can be expressed as |1-(gain characteristic of the closed loop)| 2 , or |1-(closed loop transfer function)| 2 The closed loop transfer function is expressed as G(jω)=A(ω)×e- from the gain A(ω) and phase lag θ(ω) of the Bode diagram. jθ(ω) Here, the closed loop is a speed feedback loop consisting of the subtractor 110, the speed control section 120, the filter 130, the current control section 140, and the motor 150.
[0099] (Modification) In addition to the configuration shown in FIG. 11, the control system can also be modified as follows. <Modification in which the control assistance unit is connected to the servo control unit via a network> FIG. 12 is a block diagram showing a modified example of the control system. The control system 10B shown in FIG. 12 can be applied to the control systems 10 and 10A of the first and second embodiments shown in FIG. 1 and FIG. 11. The control system 10B differs from the 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 assistance units 400-1 to 400-n via a network 500, and each of the control assistance units includes a frequency generation unit 200 and a frequency characteristic measurement unit 300. The control assistance units 400-1 to 400-n have the same configuration as the control assistance unit 400 shown in FIG. 1. The servo control units 100-1 to 100-n correspond to servo control devices, and the control assistance units 400-1 to 400-n correspond to control assistance devices, respectively. Of course, either or both of the frequency generating section 200 and the frequency characteristic measuring section 300 may be provided outside the servo control sections 100-1 to 100-n.
[0100] Here, the servo control unit 100-1 and the control assistance unit 400-1 are paired one-to-one and connected to be able to communicate with each other. The servo control units 100-2 to 100-n and the control assistance units 400-2 to 400-n are also connected in the same manner as the servo control unit 100-1 and the control assistance unit 400-1. In FIG. 12, the n pairs of the servo control units 100-1 to 100-n and the control assistance units 400-1 to 400-n are connected via a network 500, but the servo control unit and the control assistance unit of each pair of the n pairs of the servo control units 100-1 to 100-n and the control assistance units 400-1 to 400-n may be directly connected via a connection interface. These n pairs of the servo control units 100-1 to 100-n and the control assistance units 400-1 to 400-n may be installed in the same factory, for example, or may be installed in different factories.
[0101] The network 500 may be, for example, a local area network (LAN) constructed in a factory, the Internet, a public telephone network, or a combination of these. There are no particular limitations on the specific communication method in the network 500, and whether the connection is wired or wireless.
[0102] <Flexibility in system configuration> In the above-described embodiment, the servo control units 100-1 to 100-n and the control assistance units 400-1 to 400-n are each paired one-to-one and connected to each other so as to be able to communicate with each other. However, for example, one control assistance unit may be connected to multiple servo control units via a network 500 so as to be able to communicate with each other and provide control assistance to each servo control unit. In this case, the functions of one control support unit may be distributed among multiple servers as appropriate to form a distributed processing system. Also, the functions of one control support unit may utilize virtual server functions on the cloud.
[0103] In addition, when there are n control assistance units 400-1 to 400-n corresponding to n servo control units 100-1 to 100-n of the same model name, the same specifications, or the same series, the estimation results in each control assistance unit 400-1 to 400-n may be shared. By doing so, it becomes possible to construct a more optimal model.
[0104] The first and second embodiments and their modifications have been described above. Each component included in the control system of each embodiment and modification can be realized by hardware, software, or a combination of these. In addition, the servo control method performed by the cooperation of each component included in the above control system can also be realized by hardware, software, or a combination of these. Here, "realized by software" means that it is realized by a computer reading and executing a program.
[0105] The program can be stored and provided to a computer using various types of non-transitory computer readable media. The non-transitory computer readable media includes various types of tangible storage media. The non-transitory computer readable media is, for example, a magnetic recording medium (e.g., a hard disk drive), a magneto-optical recording medium (e.g., a magneto-optical disk), a CD-ROM (Read Only Memory), a CD-R, a CD-R / W, a semiconductor memory (e.g., a mask ROM, a PROM (Programmable ROM), an EPROM (Erasable PROM), a flash ROM, or a RAM (random access memory)).
[0106] Although each of the above-described embodiments is a preferred embodiment of the present invention, the scope of the present invention is not limited to only the above-described embodiments, and the present invention can be implemented in various modified forms without departing from the gist of the present invention.
[0107] The control assistance device, the control system, and the filter adjustment method according to the present disclosure can take various forms including the above-described embodiment and having the following configurations. (1) A control assistance device (e.g., a control assistance unit 400) that provides assistance for setting a first plurality of filters (e.g., filters 130-1 to 130-m) provided in a servo control device (e.g., a servo control unit 100) that controls a motor (e.g., a motor 150), A resonance detection unit (e.g., resonance detection unit 401) that detects a plurality of resonance points in the frequency characteristics of the input / output gain of the servo control device, the resonance points being measured based on an input signal and an output signal whose frequency changes; a filter setting unit (for example, the filter setting unit 402) that sets a second plurality of filters, the number of which exceeds the first plurality of filters, in order to suppress the plurality of resonance points; a grouping unit (e.g., the grouping unit 404) that groups the second plurality of filters to set the first plurality of filters; A control assistance device comprising:
[0108] According to this control assistance device, it is possible to determine the filter allocation so as to satisfy the constraint on the number of filters while realizing resonance suppression.
[0109] (2) The control assistance device described in (1) above, wherein the grouping unit includes a machine learning unit (e.g., machine learning unit 420) that determines grouping candidates having at least one composite filter that combines two or more filters of the second plurality of filters, and a grouping determination unit (e.g., grouping determination unit 410) that determines a combination of the first plurality of filters from the plurality of grouping candidates determined by the machine learning unit.
[0110] (3) The control assistance device described in (1) above, wherein the grouping unit determines a combination of the first plurality of filters by providing at least one composite filter that combines two or more filters of the second plurality of filters based on a predetermined rule.
[0111] (4) The control assistance device according to (1) above, wherein the grouping unit groups the second plurality of filters one by one to determine a combination of the first plurality of filters.
[0112] (5) The control assistance device described in (1) above, wherein the grouping unit provides at least one composite filter that combines two or more filters of the second plurality of filters, searches for all patterns for setting the first plurality of filters, and determines a combination of the first plurality of filters from among all the patterns.
[0113] (6) A control assistance device according to any one of (1) to (5) above, wherein the filters of the first plurality of filters are notch filters and / or low-pass filters.
[0114] (7) A control assistance device according to any one of (1) to (6) above, wherein even when the number of the second plurality of filters is equal to or less than a constraint number, the grouping unit groups the second plurality of filters to set the first plurality of filters.
[0115] (8) A servo control device (e.g., servo control unit 100) that controls a motor (e.g., motor 150); A control assistance device (e.g., a control assistance unit 400) according to any one of (1) to (7) above, which detects a plurality of resonance points in a frequency characteristic of an input / output gain and an input / output phase delay of the servo control device, sets a number of second filters that exceeds a number of first filters set in the servo control device in order to suppress the plurality of resonance points, and sets the first filters by grouping the second filters; A control system comprising:
[0116] According to this control system, it is possible to determine the filter allocation so as to satisfy the constraint on the number of filters while realizing resonance suppression.
[0117] (9) a frequency generator for generating a signal having a variable frequency and inputting the signal to the servo control device; a frequency characteristic measuring device for measuring frequency characteristics of an input / output gain and a phase delay of the servo control device based on the signal and an output signal of the servo control device; The control system according to (8) above, comprising:
[0118] (10) The servo control device includes a current feedback loop that controls a current flowing through the motor, and a feedback loop that includes the current feedback loop and has the filter, a frequency generator that generates a first signal having a varying frequency and inputs the first signal into the current feedback loop; a frequency characteristic measuring unit that measures frequency characteristics of an input / output gain and a phase delay of the servo control device based on the first signal and a second signal input to the current feedback loop in the feedback loop; The control system according to claim 8 or 9, further comprising:
[0119] (11) A filter adjustment support method for a control support device (e.g., a control support unit 400) that provides support for setting a first plurality of filters provided in a servo control device (e.g., a servo control unit 100) that controls a motor (e.g., a motor 150), comprising: Detecting a plurality of resonance points in a frequency characteristic of an input / output gain of the servo control device, the frequency characteristic being measured based on an input signal and an output signal having a changing frequency; setting a second plurality of filters in excess of the number of the first plurality of filters to suppress the plurality of resonance points; The filter adjustment method includes grouping the second plurality of filters to set the first plurality of filters.
[0120] According to this filter adjustment method, it is possible to determine the filter allocation so as to satisfy the constraint on the number of filters while suppressing resonance. [Explanation of symbols]
[0121] 10, 10A, 10B Control System 100, 100-1 to 100-n Servo control unit 110 Subtractor 120 Speed control section 130, 130-1~130-m Filter 140 Current control section 150 Motor 200 Frequency Generation Unit 300 Frequency characteristic measurement section 400, 400-1~400-n Control Support Department 401 Resonance detection unit 402 Filter setting section 403 Frequency characteristic estimation unit 404 Grouping Section 410 Grouping Decision Unit 420 Machine Learning Department 421 Status information acquisition unit 422 Learning Department 423 Behavioral Information Output Unit 424 Value Function Memory Unit 425 Grouping candidate output section 500 Network
Claims
1. A control assistance device that provides assistance for setting a first plurality of filters provided in a servo control device that controls a motor, comprising: a resonance detection unit that detects a plurality of resonance points in a frequency characteristic of an input / output gain of the servo control device, the frequency characteristic being measured based on an input signal and an output signal whose frequency changes; a filter setting unit that sets a second plurality of filters, the number of which exceeds the first plurality of filters, in order to suppress the plurality of resonance points; a grouping unit that groups the second plurality of filters to set the first plurality of filters; A control assistance device comprising:
2. the grouping unit includes a machine learning unit that determines grouping candidates including at least one composite filter that combines two or more filters of the second plurality of filters; and a grouping determination unit that determines a combination of the first plurality of filters from the grouping candidates determined by the machine learning unit; The machine learning unit is a state information acquisition unit that acquires state information including an input / output gain and an input / output phase delay obtained by driving the servo control device based on a filter coefficient set in a filter combination including one of the composite filters; a behavior information output unit that outputs behavior information including adjustment information for the coefficient of the filter; a reward output unit that outputs a value of a reward in reinforcement learning based on the input-output gain and the input-output phase delay output from the state information acquisition unit; a value function update unit that updates a value function based on the reward value output by the reward output unit, the state information, and the action information; a grouping candidate output unit that calculates coefficients of each filter in a combination of filters including one or more composite filters that maximizes the value based on the updated value function, and sets the coefficients of the combination of filters as grouping candidates; 2. The control assistance device according to claim 1, wherein the grouping determination unit obtains an input / output phase delay by driving the servo control device with a combination of filters of the grouping candidates for each of the plurality of grouping candidates, and determines the first combination of the plurality of filters from the plurality of grouping candidates based on the obtained plurality of input / output phase delay characteristics.
3. 2. The control assistance device according to claim 1, wherein the grouping unit determines a combination of the first plurality of filters by providing at least one composite filter that combines two or more filters of the second plurality of filters based on a predetermined rule.
4. 2. The control assistance device according to claim 1, wherein the grouping unit determines a combination of the first plurality of filters from the second plurality of filters by repeatedly operating the servo control device to adjust a coefficient of the composite filter each time a composite filter is provided by grouping adjacent filters in the second plurality of filters.
5. 2. The control assistance device according to claim 1, wherein the grouping unit provides at least one composite filter that combines two or more of the second plurality of filters, searches for all patterns for setting the first plurality of filters, and determines a combination of the first plurality of filters from among all the patterns.
6. A control assistance device described in any one of claims 1 to 5, wherein the filters of the first plurality of filters are notch filters and / or low-pass filters.
7. a servo control device for controlling the motor; A control assistance device according to any one of claims 1 to 6, which detects a plurality of resonance points in a frequency characteristic of an input / output gain and an input / output phase delay of the servo control device, and sets a number of second filters that exceeds a number of first filters set in the servo control device in order to suppress the plurality of resonance points, and sets the first filters by grouping the second filters; A control system comprising:
8. a frequency generator for generating a signal having a variable frequency and inputting the signal to the servo controller; a frequency characteristic measuring device for measuring frequency characteristics of an input / output gain and a phase delay of the servo control device based on the signal and an output signal of the servo control device; The control system of claim 7, comprising:
9. The servo control device includes a current feedback loop that controls a current flowing through the motor, and a feedback loop that includes the current feedback loop and has the first plurality of filters; a frequency generator that generates a first signal having a varying frequency and inputs the first signal into the current feedback loop; a frequency characteristic measuring unit that measures frequency characteristics of an input / output gain and a phase delay of the servo control device based on the first signal and a second signal input to the current feedback loop in the feedback loop; 9. A control system according to claim 7 or 8, comprising:
10. A filter adjustment support method for a control support device that provides support for setting a first plurality of filters provided in a servo control device that controls a motor, comprising: Detecting a plurality of resonance points in a frequency characteristic of an input / output gain of the servo control device, the frequency characteristic being measured based on an input signal and an output signal having a changing frequency; setting a second plurality of filters in excess of the number of the first plurality of filters in order to suppress the plurality of resonance points; The filter adjustment support method includes grouping the second plurality of filters to set the first plurality of filters.
Citation Information
Patent Citations
Controller for stage
JP1993035335A
Motor controller
JP2006288124A
Motor control apparatus
JP2009296746A
Machine learning apparatus, control device and machine learning method
JP2020057211A
