Adaptive control system and adaptive control method

Abstract

To provide an adaptive control system and an adaptive control method, capable of achieving both realization of adaptive control with high responsibility and easy gain adjustment.SOLUTION: An adaptive control system includes: a control circuit; and a parallel feed-forward compensation circuit that outputs a compensation value for compensating a control value fed back to the control circuit. The control circuit outputs an operation value based on the control value, the compensation value, and a command value. The parallel feed-forward compensation circuit includes: an identifier that sequentially estimates frequency response characteristics of a control target; and an adjuster that adjusts the compensation value based on the frequency response characteristics. In the control circuit, a plurality of adaptive gains are adjusted in a manner such that a normative model so designed as to provide a predetermined response is followed by the control value outputted by the control target. The plurality of adaptive gains include: a feed-forward gain that is provided to deviation for the output of the normative model of the command value; and a feed-back gain that is provided to deviation for the control value and the compensation value of the output of the normative model.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to adaptive control systems and methods. [Background technology]

[0002] There is known an adaptive control method for estimating parameters of a control target whose parameters are unknown while stabilizing a control system. In this type of adaptive control, the following Patent Document 1 has been proposed as an adaptive control device that can automatically and easily perform optimal adaptive control while preventing deterioration of responsiveness. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2013-50767 A Summary of the Invention [Problem to be solved by the invention]

[0004] The adaptive control device of Patent Document 1 above has room for improvement in terms of shortening the adjustment time of the adaptive gain.

[0005] The present disclosure has been made to solve the above-mentioned problems, and aims to provide an adaptive control system and an adaptive control method that can achieve both highly responsive adaptive control and easy gain adjustment. [Means for solving the problem]

[0006] An adaptive control system according to one aspect of the present disclosure includes a control circuit that outputs an operating value to a controlled object, and a parallel feedforward compensation circuit that outputs a compensation value based on the operating value for compensating for a control value output from the controlled object and fed back to the control circuit, the control circuit outputs the operating value based on the control value, the compensation value, and a command value, the parallel feedforward compensation circuit includes an identifier that sequentially estimates a frequency response characteristic of the controlled object, and an adjuster that adjusts the compensation value based on the frequency response characteristic, the control circuit adjusts a plurality of adaptive gains such that the control value output by the controlled object follows a reference model designed to give a predetermined response, and the plurality of adaptive gains include a feedforward gain that is applied to a deviation of the command value from an output of the reference model, and a feedback gain that is applied to a deviation of the output of the reference model from the control value and the compensation value.

[0007] Moreover, an adaptive control method according to another aspect of the present disclosure is an adaptive control method using a control system configured by adding a parallel feedforward compensation circuit to a control object, the method including: outputting an operating value to the control object; outputting a compensation value for compensating for a control value output from the control object and fed back based on the operating value; outputting the operating value based on the control value, the compensation value, and a command value; sequentially estimating a frequency response characteristic of the control object to output the compensation value, and adjusting the compensation value based on the frequency response characteristic; and adjusting a plurality of adaptive gains when outputting the operating value so that the control value output by the control object follows a reference model designed to give a predetermined response, the plurality of adaptive gains including a feedforward gain for a deviation between the command value and an output of the reference model, and a feedback gain applied to a deviation of the output of the reference model from the control value and the compensation value. Effect of the Invention

[0008] The present disclosure can achieve both highly responsive adaptive control and easy gain adjustment. [Brief description of the drawings]

[0009] [Figure 1] FIG. 1 is a block diagram illustrating a schematic configuration example of an adaptive control system according to an embodiment of the present disclosure. [Diagram 2] FIG. 2 is a graph of an open-loop response including a PFC to explain the effect of a PFC in a control device using a typical PFC. [Diagram 3] FIG. 3 is a flowchart showing a flow of PFC adjustment in the adaptive control system shown in FIG. 1 in this embodiment. [Figure 4] FIG. 4 is a graph showing the frequency response characteristics of a controlled object and the frequency response characteristics of a PFC designed accordingly. [Diagram 5] FIG. 5 is a graph showing the frequency response characteristics of an expanded control system based on the frequency response characteristics of the controlled object and PFC shown in FIG. [Figure 6] FIG. 6 is a block diagram showing a schematic configuration example when a dynamic compensation circuit is added to the adaptive control system shown in FIG. [Figure 7] FIG. 7 is a block diagram showing a schematic configuration example of an adaptive control system including a SAC circuit in a conventional configuration. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. In the following, the same or corresponding elements are denoted by the same reference numerals throughout the drawings, and duplicated descriptions thereof will be omitted.

[0011] [Overall configuration] 1 is a block diagram showing a schematic configuration example of an adaptive control system according to an embodiment of the present disclosure. As shown in FIG. 1, the adaptive control system 1 of the present embodiment includes a control circuit 3 that outputs an operation value u(t) to a controlled object 2, and a compensation value y(t) for compensating for a control value y(t) output from the controlled object 2 and fed back to the control circuit 3. fand a parallel feed-forward compensation circuit 4 that outputs a control signal u(t) based on the control signal u(t). Hereinafter, the parallel feed-forward compensation circuit is also abbreviated as PFC (Parallel Feed-forward Compensator).

[0012] The controlled object 2 in this embodiment has an integral transfer function in which the control value y(t) output for a constant input operation value u(t) rises in a ramp-like manner. For example, the controlled object 2 having an integral transfer function includes a hydraulic pump used in a press machine, a pressure test machine, etc. in which the discharge pressure rises in a ramp-like manner for a constant hydraulic oil flow rate. For example, in the case of a motor-driven hydraulic pump, the rotation speed of the motor can be the operation value u(t) and the discharge pressure can be the controlled value y(t). The adaptive control system 1 of the present disclosure is applicable not only to a controlled object 2 having an integral transfer function, but also to a controlled object 2 whose transfer function can be approximated to an integral system.

[0013] The control circuit 3 calculates the control value y(t) output from the controlled object 2 and the compensation value y(t) output from the PFC 4. f The control circuit 3 and the PFC 4 output an operation value u(t) based on a command value r(t) and a command value r(t). The control circuit 3 and the PFC 4 each include a processor and a memory. The memory stores a calculation program for performing a predetermined digital calculation and various data. The processor includes a computer such as a microcontroller, a personal computer, or a PLC (Programmable Logic Controller). For example, the processor includes a processing circuit such as a CPU or an MPU, a register, a RAM, and peripheral circuits.

[0014] The control circuit 3 and PFC4 may, for example, be stored in a server such as a cloud server, and may execute control operations while communicating with target devices including the control target 2 based on an adaptive control program stored in the server.

[0015] It should be noted that the functions of the elements disclosed herein can be performed using circuits or processing circuits including general purpose processors, special purpose processors, integrated circuits, ASICs (Application Specific Integrated Circuits), conventional circuits, or combinations thereof configured or programmed to perform the disclosed functions. A processor is considered a processing circuit or circuit because it includes transistors and other circuits. In this specification, a circuit, unit, or means is hardware that performs the recited functions or hardware that is programmed to perform the recited functions. The hardware may be hardware disclosed in this specification or other known hardware that is programmed or configured to perform the recited functions. In the case where the hardware is a processor that is considered a type of circuit, the circuit, unit, or means is a combination of hardware and software, and the software is used to configure the hardware or processor.

[0016] The PFC 4 calculates a compensation value y based on the operation value u(t) output from the control circuit 3. f (t), an identifier 6 which sequentially identifies a model of the controlled object 2 to estimate a transfer function of the controlled object 2, and an identifier 6 which estimates a frequency response characteristic of the controlled object 2 based on the transfer function identified by the identifier 6 and outputs a compensation value y f and an adjuster 7 for adjusting (t).

[0017] Fig. 2 is a graph of an open loop response including PFC to explain the effect of PFC in a control device using a general PFC. However, the PFC in Fig. 2 does not include the identifier 6 and regulator 7 shown in Fig. 1. As shown in Fig. 2, in general, the control value y output from the controlled object 2 has a response delay with respect to the operation value u(t) of the control circuit 3. In response to this, a compensation value y is generated by the PFC to compensate for the response delay of the controlled object 2. fThis generates a pseudo output based on the control value y output from the controlled object 2 and the compensation value y output from the PFC. f There is no response delay in the output of the expanded control system that combines (t) with PFC. Since response delay is the main cause of instability in feedback control, compensating for the response delay with PFC ensures basic stability and has the effect of greatly simplifying the design of the control circuit 3. A specific example of PFC is the transfer function G f Some have (s).

[0018]

number

[0019] The control value y(t) is the PFC compensation value y f In order to remove the offset caused by adding (t), the PFC may have a low-frequency cutoff characteristic as follows:

[0020]

number

[0021] The compensation value y output from the PFC f If (t) is large, the control system becomes more stable, but the compensation value y f If (t) is made too large, the output of the augmented control system will deviate from the control value y(t) output from the controlled object 2, resulting in poor responsiveness.

[0022] In contrast, according to the present embodiment, the compensation value y f (t) is automatically adjusted, so the compensation value y f (t) does not need to be large, and this prevents deterioration of responsiveness. Furthermore, unlike the conventional automatic PFC adjustment method, the compensation value y fSince there is no need to readjust (t) and the control parameters are adjusted from the frequency response characteristics, the tolerance to modeling errors is greater than in the conventional configuration in which identified parameters are directly used as control parameters. In other words, even if the modeling error is somewhat large, the control parameters can be appropriately adjusted as long as the tendency of the frequency response characteristics can be grasped. Therefore, with the above configuration, optimal adaptive control can be performed automatically and simply while preventing deterioration of responsiveness.

[0023] [How to adjust PFC] Hereinafter, the compensation value y f A method of adjusting the control value d(t) will be described. FIG. 3 is a flowchart showing the flow of adjustment of the PFC in the adaptive control system shown in FIG. 1 in this embodiment. As shown in FIG. 3, an operating value u(t) which is input data of the controlled object 2 and a control value y(t) which is output data of the controlled object 2 are input to the identifier 6 of the PFC 4. The input operating value u(t) and control value y(t) are applied to a band-pass filter to remove components outside a predetermined frequency range such as noise components (step S1). The band-pass filter includes a high-pass filter and a low-pass filter. The operating value d(t) after filtering is u and the filtered control value d y is resampled (step S2).

[0024] Thereafter, the identifier 6 performs recursive identification using the resampled values ​​(step S3). In this embodiment, the identifier 6 estimates the frequency response characteristic of the control target 2 by recursively identifying the model of the control target 2 and determining the transfer function of the control target 2. At this time, the identifier 6 performs identification by applying a linear black box model. As the linear black box model, a model called an ARX model is preferably used. This makes it possible to estimate the above-mentioned frequency response characteristic by utilizing a known recursive identification method. In addition, the identifiable control target 2 is not limited to a specific model, and can be applied to various control targets 2, making it possible to configure a highly versatile adaptive control system. Specifically, the model of the control target 2 is described as follows.

[0025]

number

[0026] Here, u r (k) denotes the operation value at time k after resampling, and y r (k) denotes the control value at time k after resampling, v(k) denotes the disturbance term, km denotes the dead time, z denotes the time shift operator for one sample, and z[x(k)]=x(k+1) holds.

[0027] Also, A(z -1 ) and B(z -1 ) can be expressed as follows:

[0028]

number

[0029] Here, a 1 ,a 2 ,…,a na and b 1 ,b 2 ,…,b nb indicates the denominator and numerator parameters to be estimated, and na, nb indicate the numbers of denominator and numerator parameters of the identification model, respectively.

[0030] At this time, the output data y at time k based on the input and output data up to time k-1 is r (k) one-step-ahead forecast value y p (k) can be expressed as follows:

[0031]

number

[0032] Here, θ denotes a parameter vector, and φ(k) denotes a data vector at time k.

[0033] In this case, if the probabilistic fluctuation of the parameter vector θ indicates the fluctuation of the control target 2, it can be expressed as follows.

[0034]

number

[0035] Here, Q indicates the parameter variance, i.e., the fluctuation range, and R indicates the variance of the observation noise. Note that the parameter variance Q is set to 0 in the steady state, i.e., in a state where there is no change in the input and output. The parameter variance Q and the observation noise variance R are the design parameters of the PFC4.

[0036] In this embodiment, the identifier 6 estimates each coefficient in the polynomial expression that is a parameter of the linear black-box model by using a Kalman filter. That is, the identifier 6 applies the Kalman filter based on the above formula (6) to estimate the parameter vector θ.

[0037] The estimation procedure using the Kalman filter is specifically described below. First, the initial value θ i (k) and the initial value of the error covariance matrix P i By giving (k), the identifier 6 can obtain the prediction error ε i (k) and the Kalman gain W(k).

[0038]

number

[0039] From the above equations (7) and (8), the parameter estimate θ(k) and the error covariance matrix P(k) are corrected as follows:

[0040]

number

[0041] Furthermore, the time step is updated to obtain the initial value θ i (k+1) and the initial value of the error covariance matrix P i Calculate (k+1).

[0042]

number

[0043] However, in the steady state, the parameter variance Q = 0, so the initial value P i (k+1) is only P(k).

[0044] In this manner, the parameter vector θ is estimated successively.

[0045] Here, the transfer function G(z) of the controlled object 2 is expressed as follows.

[0046]

number

[0047] The above formula (13) can be expressed by the parameter vector θ estimated by the identifier 6. As described above, by applying a Kalman filter to a linear black-box model, it is possible to estimate the frequency response characteristics of the control target 2 using a known configuration.

[0048] Next, the adjuster 7 designs the PFC4 based on the estimated transfer function G(z) of the controlled object 2. In this embodiment, the PFC4 is a first-order lag system represented by equation (1), and the adjuster 7 multiplies the frequency and gain at which the phase lag of the controlled object 2 is equal to or greater than a predetermined value by predetermined coefficients to determine the break point frequency ω f and gain K f In the following, we design the corner frequency ω f is the PFC frequency ω f Also, the gain K f is the PFC gain Kf Specifically, the regulator 7 first determines whether the phase delay of the controlled object 2 is φ by using the transfer function G(z) of the controlled object 2 that has been identified. p The frequency ω p is calculated by numerical search (step S4). p When p =|G(z=exp(jω p T))| is calculated (step S5), where T indicates the control period.

[0049] The regulator 7 then adjusts the frequency ω p and gain K p (Steps S6 and S7). The smoothing filter is not particularly limited, but for example, a moving average filter is applied. When a moving average filter is used, the frequency ω pf and the post-filter gain K pf is calculated as follows:

[0050]

number

[0051] Here, ns indicates the number of data points for which the moving average is taken.

[0052] The filtered frequency ω pf and the post-filter gain K pf Using this, the regulator 7 calculates the phase lag of the controlled object 2 based on the identified transfer function G(z) of the controlled object 2 by a predetermined threshold value φ p The frequency ω p and gain K p A given coefficient (frequency coefficient α w and the gain factor α k ) to obtain the transfer function G f PFC frequency ω of (z) f and PFC gain K f is designed as follows (steps S8 and S9):

[0053]

number

[0054] Here, the frequency coefficient α w and the gain factor α k is a design parameter.

[0055] The PFC frequency ω obtained from the above f and PFC gain K f The transfer function G of the PFC calculator 5 is f (s) is calculated (step S10). The calculated transfer function G f (s) based on the compensation value y f In this embodiment, the regulator 7 adjusts the PFC frequency ω f and PFC gain K f It is determined whether the value of exceeds a predetermined upper limit, and if it does, a limiter is applied so that it does not exceed the upper limit (steps S11 and S12). f This effectively prevents (s) from going outside the adjustment range.

[0056] By the way, the transfer function G of PFC4 f When (s) is calculated by the PFC calculator 5, the continuous-time transfer function G f The discrete-time transfer function G f d (z) is used.

[0057]

number

[0058] In addition, d f is the discrete-time transfer function G of the PFC4 f d (z) is the direct term in the discrete-time transfer function G fd (z) means the transfer function of PFC4 excluding the direct term. At this time, the compensation value y f (t) is calculated as follows:

[0059]

number

[0060] Here, Y f d (k) means the compensation value excluding the direct term. Note that equation (18) can also be expressed as follows:

[0061]

number

[0062] Thus, the compensation value y f By adjusting (t), the compensation value y output from the PFC 4 can be effectively output for various controlled objects 2 with a simple configuration. f (t) can be adjusted.

[0063] [SAC circuit] Next, the control circuit 3 in this embodiment will be described. As shown in Fig. 1, the control circuit 3 includes a simple adaptive control (SAC) circuit in which a plurality of adaptive gains are adjusted so that the control value y(t) output by the controlled object 2 follows a reference model designed to give a predetermined ideal response.

[0064] The multiple adaptive gains are calculated based on the reference value r(t) and the output y m (t) is the feedforward gain K r and the output of the reference model, y m (t), control value y(t) and compensation value y f (t) Feedback gain K given to the deviation from e Including,

[0065] For this purpose, the control circuit 3 includes, as a SAC circuit, a reference model applicator 31, subtractors 32, 33, 34, a first gain multiplier 35, a second gain multiplier 36, and a first adder 37. The reference model applicator 31 applies a reference model to an input command value r(t) to obtain a reference output y m The first subtractor 32 subtracts the reference output y from the command value r(t). m Subtract (t).

[0066] The second subtractor 33 subtracts the reference output y m The third subtractor 34 subtracts the compensation value y(t) from the output e(t) of the second subtractor 33. f In the configuration of Figure 1, the reference output y m (t) and subtract the control value y(t) from the result to obtain the compensation value y f (t), but first subtract the control value y(t) from the compensation value y f After adding (t), the reference output y m That is, the control circuit 3 may subtract the compensation value y(t) from the control value y(t) instead of the second subtractor 33 and the third subtractor 34. f (t) and the reference output y m and a subtractor that subtracts the output of the adder from (t).

[0067] The first gain multiplier 35 multiplies the output r a (t) is the feedforward gain K r The second gain multiplier 36 multiplies the output e of the third subtractor 34. a (t) feedback gain K e The first adder 37 multiplies the output u of the first gain multiplier 35. r (t) and the output u of the second gain multiplier 36 e (t) is added to obtain the operation value u before steady-state error compensation, which will be described later. o Calculate (t).

[0068] The reference model is expressed by the following discrete-time state equation so that it can be calculated by a computer.

[0069]

number

[0070] Here, A m ,b m ,c m ,d m indicates the parameters of the reference model.

[0071] In general, for the SAC circuit to operate properly, it is necessary for the controlled object 2 to satisfy the almost strictly positive real (ASPR) condition. However, in general, the controlled object 2 has a response delay such as dead time, and therefore often does not satisfy the ASPR condition. Therefore, in this embodiment, as described above, the output of the PFC4 is added to the output of the controlled object 2 to configure an augmented control system, and the SAC circuit is applied to the augmented control system after the augmented control system satisfies the ASPR condition.

[0072] In this embodiment, the PFC 4 has a feedback gain K e The manipulated variable component u is given e (t) based on the compensation value y f For this reason, the PFC calculator 5 is connected so as to be applied only to the second gain multiplier 36. More specifically, the PFC calculator 5 receives the output u e (t) is input, and the compensation value y f The output e(t) of the second subtractor 33 input to the third subtractor 34 is multiplied by the reference output y m (t) minus the control value y(t), and the output u of the second gain multiplier 36 e (t) is the manipulated variable u based on the feedforward gain Kr r (t) and the operation value component u w Does not include (t).

[0073] [Concepts for adjusting PFC] Here, the concept behind the adjustment method of the PFC 4 will be described. FIG. 4 is a graph showing the frequency response characteristics of a certain controlled object and the frequency response characteristics of a PFC designed accordingly. FIG. 4 shows the frequency response characteristics assuming a hydraulic device as the controlled object 2. As shown in FIG. 4, in this controlled object 2, a phase lag of about 180° or more occurs at frequencies of about 13 Hz or more. In feedback control, if there is a frequency range where the phase lag is 180° or more, there is a possibility that the control system will become unstable when feedback is performed with a gain of a certain magnitude or more.

[0074] Therefore, the output y of PFC4, whose phase delay is less than 90°, f is designed to be larger than the output y of the controlled object 2 in the frequency region where the phase delay of the controlled object 2 is 180° or more. As a result, in the frequency region where the phase delay of the controlled object 2 is 180° or more, the output y of the PFC4 with a phase delay of less than 90° is f In the example of FIG. 4, the phase delay of the control object 2 is a predetermined value (threshold value) φ p = 180°, the frequency is about 13Hz (ω p =81.2[rad / s]), and the gain of control target 2 at this time is about -49dB (K p = 0.0035), so the PFC gain when the phase lag of controlled object 2 is 180° is K f = 0.005, and the PFC frequency is ω f =100[rad / s].

[0075] FIG. 5 is a graph showing the frequency response characteristics of an expanded control system based on the frequency response characteristics of the controlled object and PFC shown in FIG. 4. As shown in FIG. 5, the output y f The output of the augmented control system when the output y of the control object 2 is added to the output y of the control object 2 is y+y fThe frequency response characteristic stops at a phase delay of less than about 90° in the frequency range where the phase delay of the control target 2 alone is 180° or more. Thus, the PFC gain K f and the PFC frequency ω f can be suitably determined from the frequency response characteristic of the control target 2. In the examples of FIGS. 4 and 5, the phase delay threshold φ p is set to 180°, but the phase delay threshold φ p may be set to 170° or 150° with more margin. The smaller the phase delay threshold φ p , the more stable the expansion control system is, but the responsiveness deteriorates. Also, the larger the PFC gain K p and the PFC frequency ω p with respect to the gain K p and the frequency ω f of the control target 2 at the threshold φ f , the more stable the expansion control system operates, but the responsiveness deteriorates.

[0076] Note that the frequency ω p and the gain K p of the control target 2 at the threshold φ p are obtained using a numerical search method within the control period T, but detailed accuracy is not required. That is, if the frequency coefficient α w and the gain coefficient α k , which are design parameters, are taken to be large enough, it is also possible to terminate the search with a tolerance range of about ±5 to ±10° with respect to the search phase, which is the threshold φ p . The number of search times converges to the above tolerance range in about 5 to 10 times using an efficient one-dimensional search method such as the golden section search method. Therefore, even if the control period T is of a general short length of about 0.002 seconds to 0.005 seconds, numerical search within the control period T is possible.

[0077] Also, when setting the phase delay threshold φ p of the transfer function G(z) within a range of, for example, 150° to 180°, the frequency coefficient α w , which is a design parameter in Equation (15),is set to about 1.0 to 5.0, and the design parameter, the gain coefficient α k is set to around 1.0 to 2.0.

[0078] [Comparison with conventional configuration] Here, the SAC circuit in this embodiment will be compared with the SAC circuit in a conventional configuration. Fig. 7 is a block diagram showing a schematic configuration example of an adaptive control system including a SAC circuit in a conventional configuration. In the conventional system 100 shown in Fig. 7, the same components as those in this embodiment are given the same reference numerals, and the description will be omitted. Note that, although only the PFC calculator 5 is shown in Fig. 7, the PFC 40 has an identifier 6 and a regulator 7 as in Fig. 1.

[0079] The control circuit 30 of the conventional system 100 includes, as an SAC circuit, two feedforward gain multipliers 51 and 52 and an adder 53 that adds the outputs from these multipliers, instead of the first subtractor 32 and the first gain multiplier 35. The first feedforward gain multiplier 51 multiplies the command value r(t) by a first feedforward gain K r1 The second feedforward gain multiplier 52 multiplies the output y m (t) is the second feedforward gain K r2 The adder 53 multiplies the output u of the first feedforward gain multiplier 51. r1 (t) and the output u of the second feedforward gain multiplier 52 r2 The output of the adder 53 is added to the output u of the second gain multiplier 36 in the first adder 37. e (t), and the result is the manipulated value u(t).

[0080] Three adaptive gains K in the conventional system 100 r1 ,K r2 ,K e When the reference model is a first-order lag system, it is adjusted according to the proportional and integral adaptive adjustment law shown in the following equation.

[0081]

number

[0082] Here, σ is the feedback gain to prevent the divergence of the feedback gain Ke, and the coefficient β 1 ,β 2 ,β 3 and the deviation e input to the feedback gain multiplier 36 a Using (t), the value is expressed by the following formula.

[0083]

number

[0084] Here, the controlled object 2 applied to this embodiment has an integral transfer function as described above. In the controlled object 2 having an integral transfer function, the operation value u(t) in the steady state can be regarded as 0. Therefore, the first feedforward gain K r1 and the second feedforward gain K r2 The relationship between K r1 =-K r2 Therefore, in the control circuit 3 of the present embodiment, the two feedforward gains K r1 ,K r2 are combined into one, and the above adaptive adjustment law in the conventional system 100 is changed to the adaptive adjustment law shown in the following equation.

[0085]

number

[0086] In this way, in the conventional system 100, the operation value component u based on the feedforward gain r To generate (t), two feedforward gains K r1 ,K r2 In contrast, in the present embodiment, the adjustment of one feedforward gain K rBy simply adjusting r (t) can be generated. This allows the feedforward gain K r Therefore, it is possible to realize adaptive control with good response and easy gain adjustment at the same time.

[0087] For example, in the control of a hydraulic system in which the controlled object 2 is a hydraulic pump, high responsiveness and high accuracy are required while the characteristics of the controlled object 2 fluctuate greatly. For this reason, the frequency of system readjustment increases when the controlled object 2 is replaced, etc., and accordingly the frequency of readjustment of the adaptive gain also increases. In such a case, with the adaptive control system 1 of this embodiment, the time required for readjustment of the adaptive gain can be reduced compared to the conventional system 100, and therefore the time required for system readjustment when the controlled object 2 is replaced, etc. can be reduced.

[0088] In addition, the PFC 40 of the conventional system 100 calculates the compensation value y(t) based on the final operation value u(t). f (t) is calculated. That is, the compensation value y f The operation value u(t) for calculating (t) is the feedforward gain K r1 ,K r2 The manipulated variable component u based on r (t) and feedback gain K e The manipulated variable component u based on e In contrast, the PFC4 in this embodiment includes both the feedback gain K e The manipulated variable component u is given e (t) based on the compensation value y f For this reason, the PFC calculator 5 is connected so as to be applied only to the second gain multiplier 36.

[0089] In this manner, in this embodiment, the feedback gain K e The manipulated variable component u based on e (t) based on the compensation value y fSince (t) is calculated, a higher response can be obtained compared to the conventional system 100.

[0090] [Compensation for steady-state error 1] In this embodiment, as described above, one feedforward gain K r The manipulated variable component u based on r In order to introduce (t), the manipulated value u(t) in the steady state is considered to be 0. However, for example, in an actual pressure system, losses occur due to fluid leakage, friction, and the like. For this reason, the loss in the control value y(t) relative to the manipulated value u(t) in the steady state may exist as a steady-state deviation. Therefore, the adaptive control system 1 in this embodiment further includes a steady-state deviation compensation circuit 38 that compensates for the manipulated value u(t) based on the manipulated value u(t) and the control value y(t).

[0091] The steady-state deviation compensation circuit 38 estimates loss characteristics such as leakage flow characteristics using a characteristic model determined based on the relationship between the operation value u(t) and the control value y(t) at that time, and calculates a steady-state deviation compensation value u(t) that compensates for the operation value u(t). w The control circuit 3 determines the output u of the first adder 37. o At (t), the steady-state deviation compensation value u w It includes a second adder 39 which adds (t).

[0092] The specific model has a gain of K fc For example, the characteristic model is a steady-state error compensation value u(t) for the control value y(t). w The relationship between the operation value u(t) and the control value y(t) is approximated by a linear equation as follows: The steady-state error compensation circuit 38 estimates the coefficients a and b in the following equation based on the operation value u(t) and the control value y(t).

[0093]

number

[0094] Here, the control value y(t) in the above equation (24) is the reference output ym In other words, the steady-state deviation compensation circuit 38 may use y(t)=y m (t) is the steady-state error compensation value u w (t) may be calculated.

[0095] In this manner, in this embodiment, by assuming that there is no steady-state error in the steady state of the controlled object 2, the necessary feedforward gain K r While combining them into one, steady-state deviation that may occur in the actual controlled object 2 is separately compensated for by the steady-state deviation compensation circuit 38. This makes it possible to shorten the gain adjustment time and realize highly accurate adaptive control according to the actual conditions of the controlled object 2.

[0096] [Compensation for steady-state error 2] As described above, in the adaptive control system 1 of this embodiment, the output of the augmented control system follows the reference model, not the output of the controlled object 2, which is originally intended to follow the reference model. In other words, a steady-state deviation due to the adoption of the augmented control system may remain in the output of the controlled object 2. In order to remove such a steady-state deviation, the adaptive control system 1 adjusts the output y m Dynamic compensation may be performed for (t).

[0097] 6 is a block diagram showing a schematic configuration example when a dynamic compensation circuit is added to the adaptive control system shown in FIG. 1. The same components as those in FIG. 1 are given the same reference numerals and the description will be omitted. In the example of FIG. 6, the control circuit 3 converts the output u of the first gain multiplier 35 into r (t) is input, and the dynamic compensation value y s (t) and a reference output y m (t) is the dynamic compensation value y s In this case, the second subtractor 33 includes a reference output y m (t) is the dynamic compensation value y s The value obtained by subtracting the control value y(t) from the value to which (t) is added is output as e(t).

[0098] The dynamic compensation circuit 41 is a PFC4 identifier 6 of the transfer function G f (s) is used to calculate the output y m Dynamic compensation value y for (t) s In other words, the output of the regulator 7 of the PFC 4 is input to the dynamic compensation circuit 41, similarly to the PFC calculator 5. According to this configuration, the output ym(t) of the reference model is compensated for using the model of the controlled object 2 identified in the PFC 4. Therefore, it is possible to appropriately remove the steady-state deviation based on the error between the output of the controlled object 2 and the output of the extended control system.

[0099] Although the embodiment of the present invention has been described above, the present invention is not limited to the above embodiment, and various improvements, changes, and modifications are possible without departing from the spirit of the present invention.

[0100] [Other embodiments] For example, in the above embodiment, the parameters of the linear black-box model are estimated using a Kalman filter as the identifier 6, but the present invention is not limited to this. For example, the parameters of the linear black-box model may be estimated using a recursive least squares (RLS) method.

[0101] Furthermore, when the physical structure of the controlled object 2 is known, a physical model of the controlled object 2 may be applied as the identifier 6. This allows a more accurate adaptive control device to be configured. In this case, the identifier 6 may be configured to estimate the unknown constants of the physical model using a Kalman filter. This allows adaptive control using a physical model to be easily realized by utilizing a known configuration.

[0102] It is also possible to identify the control target 2 without using a linear black box model. For example, an IIR filter representing the control target 2 may be obtained using an adaptive digital filter such as HARF (Hyperstable Adaptive Recursive Filter) or SHARF (Simplified HARF). Furthermore, in the adaptive control system 1 of the present disclosure, it is not always necessary to perform model identification of the control target 2, since it is sufficient to ultimately estimate the frequency response characteristics of the control target 2. In other words, the frequency response characteristics may be directly estimated in the identifier 6. As a method for directly estimating the frequency response characteristics, for example, a method such as a short-time Fourier transform and a continuous wavelet transform can be considered.

[0103] In addition, in the above embodiment, the reference model has a transfer function of a first-order lag system. However, the transfer function applied to the reference model is not limited to this, and the reference model may have other transfer functions, such as a second-order lag system or a third-order lag system.

[0104] In the above embodiment, the adaptive control system 1 includes the steady-state deviation compensation circuit 38, but the steady-state deviation compensation circuit 38 may be omitted. The adaptive control system 1 may include the dynamic compensation circuit 41 and not the steady-state deviation compensation circuit 38. Of course, as shown in FIG. 6, the adaptive control system 1 may include both the steady-state deviation compensation circuit 38 and the dynamic compensation circuit 41.

[0105] [Summary of this disclosure] [Item 1] An adaptive control system according to one aspect of the present disclosure includes a control circuit that outputs an operating value to a controlled object, and a parallel feedforward compensation circuit that outputs a compensation value based on the operating value for compensating for a control value output from the controlled object and fed back to the control circuit, the control circuit outputs the operating value based on the control value, the compensation value, and a command value, the parallel feedforward compensation circuit includes an identifier that sequentially estimates a frequency response characteristic of the controlled object, and an adjuster that adjusts the compensation value based on the frequency response characteristic, the control circuit adjusts a plurality of adaptive gains such that the control value output by the controlled object follows a reference model designed to give a predetermined response, and the plurality of adaptive gains include a feedforward gain that is applied to a deviation of the command value from an output of the reference model, and a feedback gain that is applied to a deviation of the output of the reference model from the control value and the compensation value.

[0106] According to the above configuration, while it was necessary to adjust two feedforward gains simultaneously in parallel in the conventional method, it is possible to generate an operation value component based on a feedforward gain by adjusting only one feedforward gain. This makes it possible to adjust the feedforward gain with high accuracy and in a short time. Therefore, it is possible to achieve both adaptive control with good responsiveness and easy gain adjustment.

[0107] [Item 2] In the adaptive control system of item 1, the parallel feedforward compensation circuit may calculate the compensation value based on a manipulation value component to which the feedback gain is applied. With this, since the compensation value is calculated based on a manipulation value component based on a feedback gain, it is possible to obtain a higher responsiveness than in a conventional configuration.

[0108] [Item 3] The adaptive control system of item 1 or 2 may include a steady-state deviation compensation circuit that compensates the operating value based on the operating value and the control value, and the steady-state deviation compensation circuit may determine a steady-state deviation compensation value for the operating value using a characteristic model that is determined based on the relationship between the operating value and the control value. This makes it possible to achieve highly accurate adaptive control according to the actual conditions of the controlled object while shortening the time for gain adjustment.

[0109] [Item 4] In the adaptive control system of item 3, the characteristic model may be such that a relationship between the steady-state deviation compensation value and the control value is expressed by a linear equation.

[0110] [Item 5] The adaptive control system according to any one of items 1 to 4 may include a dynamic compensation circuit that performs dynamic compensation on an output of the reference model, the identifier recursively identifies a model of the controlled object to estimate a transfer function of the controlled object, and recursively estimates a frequency response characteristic of the controlled object based on the estimated transfer function, and the dynamic compensation circuit may determine a dynamic compensation value for the output of the reference model using the transfer function estimated by the identifier. This makes it possible to appropriately remove a steady-state error based on an error between the output of the controlled object and the output of an augmented control system.

[0111] [Item 6] In any one of the adaptive control systems according to items 1 to 5, the identifier may be configured to recursively identify a model of the controlled object, estimate a transfer function of the controlled object, and recursively estimate a frequency response characteristic of the controlled object based on the estimated transfer function.

[0112] [Item 7] In the adaptive control system of item 6, the identifier may apply a linear black-box model.

[0113] [Item 8] In the adaptive control system of item 7, the identifier may estimate each coefficient in a polynomial expression of the linear black-box model using a Kalman filter.

[0114] [Item 9] In the adaptive control system according to any one of items 1 to 8, the controlled object may have an integral transfer function in which the control value outputted increases in a ramp shape with respect to the operation value inputted at a constant value.

[0115] [Item 10] An adaptive control method according to another aspect of the present disclosure is an adaptive control method using a control system configured by adding a parallel feedforward compensation circuit to a controlled object, comprising: outputting an operating value to the controlled object; outputting a compensation value for compensating for a control value output from the controlled object and fed back based on the operating value; outputting the operating value based on the control value, the compensation value, and a command value; sequentially estimating a frequency response characteristic of the controlled object to output the compensation value, and adjusting the compensation value based on the frequency response characteristic; and, when outputting the operating value, adjusting a plurality of adaptive gains so that the control value output by the controlled object follows a reference model designed to give a predetermined response, the plurality of adaptive gains including a feedforward gain for a deviation between the command value and an output of the reference model, and a feedback gain applied to a deviation of the output of the reference model from the control value and the compensation value. [Explanation of symbols]

[0116] 1. Adaptive Control Systems 2. What is controlled? 3 Control Circuit 4 Parallel Feedforward Compensation Circuit 5 Identifier 6 regulator 38 Steady-state error compensation circuit 41 Dynamic compensation circuit

Claims

1. A control circuit that outputs an operation value to a controlled object; a parallel feedforward compensation circuit that outputs a compensation value for compensating for a control value that is output from the controlled object and fed back to the control circuit based on the operation value, the control circuit outputs the operation value based on the control value, the compensation value and a command value; the parallel feedforward compensation circuit includes an identifier that sequentially estimates a frequency response characteristic of the controlled object, and an adjuster that adjusts the compensation value based on the frequency response characteristic; the control circuit adjusts a plurality of adaptive gains so that the control value output by the controlled object follows a reference model designed to give a predetermined response; An adaptive control system, wherein the plurality of adaptive gains include a feedforward gain applied to a deviation of the command value from an output of the reference model, and a feedback gain applied to a deviation of the output of the reference model from the control value and the compensation value.

2. 2. The adaptive control system of claim 1, wherein the parallel feedforward compensation circuit calculates the compensation value based on a manipulated value component to which the feedback gain is applied.

3. a steady-state deviation compensation circuit that compensates for the operation value based on the operation value and the control value; 3. The adaptive control system according to claim 1, wherein the steady-state deviation compensation circuit determines the steady-state deviation compensation value for the operation value by using a characteristic model that is determined based on a relationship between the operation value and the control value.

4. 4. The adaptive control system according to claim 3, wherein the characteristic model expresses a relationship of the steady-state deviation compensation value with respect to the control value as a linear expression.

5. a dynamic compensation circuit that performs dynamic compensation on an output of the reference model; the identifier sequentially identifies a model of the controlled object to estimate a transfer function of the controlled object, and sequentially estimates a frequency response characteristic of the controlled object based on the estimated transfer function; 3. The adaptive control system according to claim 1, wherein the dynamic compensation circuit determines a dynamic compensation value for an output of the reference model by using the transfer function estimated by the identifier.

6. 3. The adaptive control system according to claim 1, wherein the identifier recursively identifies a model of the controlled object to estimate a transfer function of the controlled object, and recursively estimates a frequency response characteristic of the controlled object based on the estimated transfer function.

7. The adaptive control system of claim 6 , wherein the identifier applies a linear black-box model.

8. The adaptive control system of claim 7 , wherein the identifier estimates each coefficient in a polynomial representation of the linear black-box model using a Kalman filter.

9. 3. The adaptive control system according to claim 1, wherein the controlled object has a transfer function of an integral system in which the control value outputted increases in a ramp shape with respect to the operation value inputted at a constant value.

10. An adaptive control method using a control system configured by adding a parallel feedforward compensation circuit to a control target, comprising: Output the operation value to the control target, outputting a compensation value for compensating for a control value that is output from the controlled object and fed back based on the operation value; outputting the operation value based on the control value, the compensation value and a command value; Sequentially estimating a frequency response characteristic of the controlled object to output the compensation value, and adjusting the compensation value based on the frequency response characteristic; adjusting a plurality of adaptive gains so that the control value output by the controlled object follows a reference model designed to give a predetermined response when outputting the operation value; An adaptive control method, wherein the plurality of adaptive gains include a feedforward gain for a deviation between the command value and an output of the reference model, and a feedback gain applied to a deviation of the output of the reference model from the control value and the compensation value.

Images