Shaping filter acquisition method, computer program, and recording medium

Abstract

To provide technology that makes it possible to acquire a sharing filter for general purposes that shapes a reference signal while considering control constraints in a feedback control system.SOLUTION: A shaping filter acquisition method includes: a reference model setting step S2 for setting a reference model Td; a controller parameter optimization step S3 for optimizing a controller parameter ρ of a controller 30 using an evaluation function JFRIT(ρ) based on FRIT including time-series data u0(t), y0(t) on an input signal u and an output signal y and the reference model Td; a reference model optimization step S5 for optimizing the reference model Td by global optimization in which the evaluation function JFRIT(ρ) is taken as a cost function; and a shaping filter calculation step S4 for calculating a shaping filter F on the basis of a controller parameter ρ* obtained by the controller parameter optimization step S3 and the reference model Td obtained by the reference model optimization step S5.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The subject matter disclosed in this specification relates to a shaping filter acquisition method, a computer program, and a recording medium.

Background Art

[0002] For example, in a process system using heat or fluid, since it is difficult to model the first principle of the system due to non-linearity and it is intuitively easy to understand parameter adjustment based on the behavior of the controlled object, PID control is widely used as feedback control. However, it is difficult to always obtain good control results with a controller using fixed PID parameters, so in many cases, the desired control performance cannot be satisfied. Therefore, when using a PID controller, not only is it adjusted at the time of design, startup, etc., but also readjusted according to changes in usage conditions, etc.

[0003] In recent years, several methods have been proposed to satisfy the desired control performance without modeling the controlled object. Among them, data-driven control is known as one of the powerful methods. Data-driven control is a method of designing a controller that achieves a target by directly using data without using a model of the controlled object.

[0004] Several variations of data-driven control are known. For example, Non-Patent Document 1 proposes FRIT (Fictitious Reference Iterative Tuning), which is one of the variations. FRIT can adjust the controller parameters based on the experimental data obtained from a single closed-loop system control experiment. Therefore, FRIT is superior in terms of costs such as time and expense compared to other methods that require repeated experiments.

[0005] Generally, PID controllers are implemented in embedded devices. Therefore, in order to introduce a controller having controller parameters obtained by data-driven control such as FRIT, a great deal of effort may be required for embedded implementation. Also, when the detailed implementation of the PID controller implemented in the embedded device is not disclosed, the application of data-driven control methods such as FRIT is usually difficult.

[0006] Therefore, Patent Document 1 proposes a filter device that outputs a desired output signal to a control target without changing the controller itself to solve the above problems. By using this filter device, it is possible to obtain desired control performance by shaping and correcting the waveform of the reference signal even when information on the detailed implementation of the controller is not available. However, when controlling using the obtained filter device, it is unclear what input signal the controller outputs. That is, it is not always guaranteed that the input signal corresponding to the shaped reference signal falls within the output range of the controller.

[0007] In response to such problems, Non-Patent Document 2 proposes a filter device that takes into account the constraints of the input signal by introducing an input limit function that shapes the reference signal so as to satisfy the constraints of the input signal.

Prior Art Documents

Patent Documents

[0008]

Patent Document 1

Non-Patent Documents

[0009]

Non-Patent Document 1

Non-Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0010] However, in the case of Non-Patent Document 2, when the initial reference model is not appropriately set, it is difficult to obtain an appropriate shaping filter, so there is a problem of low generality.

[0011] An object of the present invention is to provide a technique capable of generally obtaining a shaping filter that shapes a reference signal while considering control constraints in a feedback control system.

Means for Solving the Problems

[0012] To solve the above problems, a first aspect is a shaping filter acquisition method for acquiring a shaping filter that shapes a reference signal input to a feedback control system including a controller and a controlled object, comprising: a) preparing time-series data of an input signal input from the controller to the controlled object and an output signal output from the controlled object; b) setting a reference model of the control system; c) optimizing the controller parameters of the controller using an evaluation function based on FRIT (Fictitious Reference Iterative Tuning) including the time-series data and the reference model set in step b); d) optimizing the reference model by global optimization using the evaluation function as a cost function; and f) obtaining the shaping filter designed to satisfy control constraints based on the controller parameters obtained in step c and the reference model obtained in step d.

[0013] The second aspect is the shaping filter acquisition method of the first aspect, further including: g) a step of repeating the steps b) to d) a predetermined number of times, and the step g) includes a step of setting the reference model having the parameters obtained by the step d) in the step b).

[0014] The third aspect is the shaping filter acquisition method of the second aspect, wherein the step d) includes a step of obtaining the calculated evaluation value of the optimization, and the step f) includes a step of determining the shaping filter based on the controller parameters and the reference model when the evaluation value is the best.

[0015] The fourth aspect is the shaping filter acquisition method of the first aspect or the second aspect, wherein the controller is a PID controller.

[0016] The fifth aspect is the shaping filter acquisition method of the fourth aspect, wherein the transfer function of the controller is represented by the formula

Equation

[0017] The sixth aspect is the shaping filter acquisition method of the fourth aspect, wherein the reference model includes at least a time constant, an order, and a dead time as parameters.

[0018] The seventh aspect is the shaping filter acquisition method of the sixth aspect, wherein the reference model is represented by the formula

Equation

[0019] The eighth aspect is the shaping filter acquisition method of the seventh aspect, wherein the evaluation function based on the FRIT is

Equation

Equation

Number

[0020] The ninth aspect is the shaping filter acquisition method of the first aspect or the second aspect, wherein the global optimization is Bayesian optimization using the evaluation function based on the FRIT as a cost function.

[0021] The tenth aspect is a computer-readable computer program that causes the computer to execute the shaping filter acquisition method of the first aspect or the second aspect.

[0022] The eleventh aspect is a computer-readable recording medium on which the computer program of the tenth aspect is recorded.

Advantages of the Invention

[0023] According to the shaping filter acquisition methods of the first aspect to the ninth aspect, since the reference model is updated by global optimization, a shaping filter for shaping the reference signal input to the feedback control system while considering control constraints can be generally obtained.

Brief Description of the Drawings

[0024]

Figure 1

Figure 2

Figure 3

Modes for Carrying Out the Invention

[0025] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. Note that the components described in this embodiment are merely examples, and are not intended to limit the scope of the present invention thereto. In the drawings, for ease of understanding, the dimensions and numbers of each part may be exaggerated or simplified as necessary.

[0026] <1. Embodiment> FIG. 1 is a block diagram showing the hardware configuration of an information processing apparatus 1 according to an embodiment. The information processing apparatus 1 is configured by a general-purpose computer installed with a dedicated computer program P. The information processing apparatus 1 includes a processor 11, a memory 12, a storage 13, an operation device 15, a display 16, and an input / output interface 17.

[0027] The processor 11 includes, for example, a CPU. The memory 12 includes, for example, a RAM which is a semiconductor memory. The storage device 13 is an auxiliary storage device and includes, for example, a hard disk drive (HDD) or a solid state drive (SSD). The storage device 13 stores the computer program P and various data.

[0028] The processor 11 executes processing according to the computer program stored in the storage device 13. The memory 12 is used as a working area for the processor 11. The operation device 15 inputs a user's operation input to the processor 11. The operation device 15 includes, for example, a keyboard or a pointing device. The display 16 displays various information based on the control of the processor 11. The display 16 is, for example, a liquid crystal display.

[0029] The input / output interface 17 is an interface for inputting data from the outside into the information processing apparatus 1 and for outputting data from the information processing apparatus 1 to the outside. The input / output interface 17 is, for example, a USB interface or the like.

[0030] The computer program P may be stored in a non-transitory recording medium such as a semiconductor memory such as a USB memory, or an optical or magnetic medium. And the computer program P of the recording medium may be provided to the information processing apparatus 1 via the input / output interface 17. Also, the computer program P may be provided to the information processing apparatus 1 via a network such as the Internet.

[0031] FIG. 2 is a block diagram showing a feedback control system 100. The feedback control system 100 includes a subtractor 20, a controller 30, a controlled object 40, and a filter device 50. The controller 30 is an implemented feedback controller, for example, a PID controller. The controller 30 performs control so that the feedback control system 100 is stabilized. The filter device 50 inputs the shaped reference signal r obtained by shaping the reference signal r input to the feedback control system 100 by the shaping filter F * to the subtractor 20. The subtractor 20 inputs the deviation E (= r * −y) between the shaped reference signal r * and the output signal y output from the controlled object 40 to the controller 30. The controller 30 outputs an input signal u (= C(ρ0)·E) to the controlled object 40 with respect to the deviation E according to the transfer function C(ρ0). The output signal y from the controlled object 50 is measured by a measuring instrument (not shown) and input to the subtractor 20.

[0032] Here, let the desired transfer function from the reference signal r to the output signal y be T d . When the controller parameters of the controller 30 are insufficient, it is difficult to realize the target response y d = T d r. Therefore, it is necessary to adjust the controller parameters so that the closed-loop system from the reference signal r to the output signal y approaches the desired transfer function T d . However, it may be difficult to easily tune the controller 30. Therefore, in the feedback control system 100, the filter device 50 shapes the reference signal r (reference signal) to obtain the shaped reference signal r *By inputting it into the subtractor 20, the desired control performance is achieved.

[0033] The information processing apparatus 1 performs a process of acquiring the shaping filter F of the filter apparatus 50. FIG. 3 is a diagram showing a flow of a process in which the information processing apparatus 1 acquires the shaping filter F.

[0034] In the following description, a case where the control target 50 is an electric needle valve whose fluid flow rate can be adjusted and the controller 30 performs PID control on the flow rate will be described. The electric needle valve is such that the opening degree of the valve can be controlled by a stepping motor, and the controller 30 adjusts the opening degree so that the deviation between the flow rate which is the output signal y and the target flow rate which is the reference signal r becomes zero.

[0035] The information processing apparatus 1 first performs an initial experiment data acquisition step S1. In the initial experiment data acquisition step S1, in an actual machine, using the already mounted controller 30 (C(ρ0)), flow rate control is performed with the reference signal r (target flow rate time series data) as an input, and a set of time series data u0(t), y0(t) which is a combination of the input signal u and the output signal y (flow rate) is acquired. The transfer function C(ρ) of the controller 30 is represented by, for example, the following equation.

[0036]

Equation

[0037] In the initial experiment data acquisition step S1, u0(t), y0(t) may be input to the information processing apparatus 1 via a recording medium such as a USB memory or a network. Then, the processor 11 of the information processing apparatus 1 may store the input u0(t), y0(t) in the memory 12 or the storage device 13.

[0038] Subsequently, the information processing apparatus 1 sets the reference model T d (reference model setting step S2). The reference model Td is represented by, for example, the following equation.

[0039] [Number]

[0040] As shown in Equation (2), the reference model T of this example d shall include, as parameters, the time constant τ, the order n, and the dead time L.

[0041] Subsequently, the information processing apparatus 1 optimizes the controller parameter ρ of the controller 30 by using u0(t) and y0(t) obtained in the initial experiment data acquisition step S1 (controller parameter optimization step S3). In the controller parameter optimization step S3, an evaluation function J FRIT (ρ) defined based on the algorithm of FRIT (Fictitious Reference Iterative Tuning) is used. The evaluation function J FRIT (ρ) is expressed, for example, by the following equation.

[0042] [Number] [Number]

[0043] As shown in Equation (3), the evaluation function J FRIT (ρ) is a function of the control parameter ρ. Therefore, in the controller parameter optimization step S3, the information processing apparatus 1 uses a non-linear programming optimization method to minimize the evaluation function J FRIT (ρ) to calculate the controller parameter ρ * . The information processing apparatus 1 stores the calculated controller parameter ρ * in the memory 12 or the storage device 13.

[0044] Subsequently, the information processing apparatus 1 calculates a shaping filter F (shaping filter calculation step S4). As a method for calculating the shaping filter F, for example, the method described in Non-Patent Document 2 can be used. Specifically, the shaping filter F includes transfer functions F1, F2 represented by the following expressions, and an input limitation function. The transfer functions F1, F2 are represented by the following expressions.

[0045]

Equation

Equation

[0046] Further, the input limitation function is represented by the following expression.

[0047]

Equation

Equation

[0048] u shown in Equation (7) p is the output of the transfer function F1. Also, u shown in Equation (8) ref is the reference input signal. u max is the maximum value of the input signal u, and u min is the minimum value of the input signal u. For example, when the filter device 50 is an electric needle valve, as a control constraint, for example, the upper limit value of the valve opening corresponds to u max and the lower limit value of the valve opening corresponds to u min As shown in Equation (8), the reference input signal u ref is designed to fall within the controllable range of the control target 40 (u min or more and u max or less).

[0049] Note that the filter device 50 calculates a shaped reference signal r * by the following equation.

Equation

[0050] The shaped reference signal r calculated according to Equation (9) * By inputting this into the closed-loop system, even when the controller 30 is not updated, an input signal u that satisfies the control constraints can be input to the controlled object 40.

[0051] The information processing apparatus 1 stores the shaping filter F calculated in the shaping filter calculation step S4 in the memory 12 or the storage device 13.

[0052] Subsequently, the information processing apparatus 1 optimizes the parameters of the reference model T d by global optimization (reference model optimization step S5). For global optimization, for example, Bayesian optimization can be used. The optimization target of Bayesian optimization is the parameters θ of the reference model T d represented by Equation (2), which are the time constant τ, the order n, and the dead time L. Also, for the cost function in Bayesian optimization, J FRIT (ρ) represented by Equation (3) is used.

[0053] Note that in the reference model optimization step S5, at the initial stage of the optimization trial of Bayesian optimization (for example, up to about 1 to 10 trials), values randomly sampled from the search range may be set for the time constant τ, the order n, and the dead time L. Also, after the initial stage, the time constant τ, the order n, and the dead time L are set based on the acquisition function of the Bayesian optimization method.

[0054] The information processing apparatus 1 stores the reference model T d obtained in the reference model optimization step S5 in the memory 12 or the storage device 13. Also, the information processing apparatus 1 stores the evaluation value (cost value) obtained in the process of Bayesian optimization in the memory 12 or the storage device 13.

[0055] After executing the reference model optimization step S5, the information processing apparatus 1 stores the number of times (execution count) the reference model optimization step S5 has been executed so far in the memory 12 or the storage device 13. Then, the information processing apparatus 1 determines whether the number of executions of the reference model update step S6 exceeds a preset default number (determination step S6).

[0056] In the determination step S6, if the execution count does not exceed the default number, the information processing apparatus 1 executes the reference model setting step S2 again. In the reference model setting step S2 in this case, a reference model T having the parameter θ optimized by the reference model optimization step S5 d is set. And based on the set reference model T d , the controller parameter optimization step S3 to the reference model optimization step S5 are executed. In this way, by repeatedly performing the reference model setting steps S2 to S5, the controller parameter ρ * and the parameters of the reference model T d are sequentially updated.

[0057] In the determination step S6, if the execution count exceeds the default number, the information processing apparatus 1 executes the shaping filter determination process S7. In the shaping filter determination process S7, the information processing apparatus 1 determines the shaping filter F (transfer functions F1, F2, and input limit function) based on the controller parameter ρ * and the reference model T d when the evaluation value (cost value) obtained by repeating the reference model optimization step S5 is the best. The shaping filter F determined in the shaping filter determination process S7 is used as the shaping filter F of the filter device 50 shown in FIG. 2.

[0058] J represented by Equation (3) FRITWhen the cost function is (ρ), the optimization of the time constant τ, the order n, and the dead time L may not be a quadratic programming problem depending on the characteristics of the controlled object 40 and the magnitude of the noise influence of the initial input data. In contrast, in this example, global search is performed by using Bayesian optimization. Therefore, it is possible to perform robust parameter search against the characteristics of the controlled object 40 and the noise influence. Therefore, it can be generally applied to the calculation of the shaping filter F.

[0059] <2. Modification example> As described above, the embodiments have been described, but the present invention is not limited to the above, and various modifications are possible.

[0060] For example, in the above embodiment, the shaping filter calculation step S4 is incorporated into the process of repeating the reference model setting step S2 to the reference model optimization step S5. However, it is not essential that the shaping filter calculation step S4 be executed during this repeated process. For example, after the repeated process, the shaping filter F may be calculated based on the controller parameters and the reference model T d that result in the best evaluation value.

[0061] Also, in the reference model optimization step S5, Bayesian optimization is used as global optimization, but other algorithms may be adopted. As global optimization, for example, an annealing method or a genetic algorithm, which is a metaheuristic method, may be used.

[0062] Also, the controlled object 40 of the feedback control system 100 is not limited to the electric needle valve.

[0063] Although the present invention has been described in detail, the above description is illustrative in all aspects, and the present invention is not limited thereto. Innumerable modifications that are not illustrated can be assumed without departing from the scope of the present invention. Each configuration described in the above embodiments and each modification can be appropriately combined or omitted as long as they do not conflict with each other.

Description of Symbols

[0064] 1: Information processing device 30: Controller 40: Controlled object 100: Feedback control system

Claims

1. A shaping filter acquisition method for acquiring a shaping filter that shapes a reference signal input to a feedback control system including a controller and a controlled object, comprising: a) preparing time series data of an input signal input from the controller to the controlled object and an output signal output from the controlled object; b) setting a reference model of the control system; c) optimizing the controller parameters of the controller using an evaluation function based on FRIT (Fictitious Reference Iterative Tuning) including the time series data and the reference model set in step b); d) optimizing the reference model by global optimization using the evaluation function as a cost function; f) obtaining the shaping filter designed to satisfy control constraints based on the controller parameters obtained in step c) and the reference model obtained in step d); A shaping filter acquisition method including the above steps.

2. The shaping filter acquisition method according to claim 1, further comprising: g) repeating steps b) to d) a predetermined number of times, wherein step g) includes setting the reference model having the parameters obtained in step d) in step b). A shaping filter acquisition method further comprising the above steps.

3. The shaping filter acquisition method according to claim 2, wherein: step d) includes obtaining the calculated evaluation value of the optimization; step f) includes determining the shaping filter based on the controller parameters and the reference model when the evaluation value is the best.

4. The shaping filter acquisition method according to claim 1 or claim 2, wherein: the controller is a PID controller.

5. The shaping filter acquisition method according to claim 4, wherein: the transfer function of the controller is expressed by the formula A shaping filter acquisition method represented by the above formula. 【Number 1】

6. The shaping filter acquisition method according to claim 4, wherein: the reference model includes at least a time constant, an order, and a dead time as parameters.

7. The shaping filter acquisition method according to claim 6, wherein: the reference model is expressed by the formula A shaping filter acquisition method represented by the above formula. 【Number 2】

8. The shaping filter acquisition method according to claim 7, wherein: ​ The evaluation function based on the FRIT is 【Number 3】 [Number 4] 【Number 2】 a shaping filter acquisition method represented by **Claim 9** A shaping filter acquisition method according to claim 1 or claim 2, wherein the global optimization is Bayesian optimization using the evaluation function based on the FRIT as a cost function. **Claim 10** A computer-readable computer program that causes the computer to execute the shaping filter acquisition method according to claim 1 or claim 2. **Claim 11** A computer-readable recording medium, on which the computer program according to claim 10 is recorded.

Images