JPS63158607A - Determining device for control constant of control system - Google Patents

Abstract

PURPOSE:To shorten the time required for determination of an optimum control constant by inferring whether specifications are satisfied or not from the pole and the zero position of a closed loop control system to adjust weight matrixes Q and R. CONSTITUTION:An equation is solved in a Riccati equation solving part 5 based on information from a dynamic characteristic input part 3 and a Q.R setting part 4. A control constant operating part 6 which receives the result obtains the control constant in accordance with a prescribed operation. A pole and zero calculating part 7 of the closed loop control system calculates the pole and the zero position of the closed loop control system on the basis of the result, and calculated results are inputted to an inference part 8. Design specifications from a control design specification setting part 10 are inputted to the inference part 8 to infer whether specifications are satisfied or not, and the method to adjust values of Q and R is inferred if they are not satisfied. A data base 9 where required and sufficient knowledge information is stored is used for this inference. Thus, the time required for determination of the optimum control constant is shortened.

Description

[Detailed Description of the Invention] [Object of the Invention] (Industrial Application Field) The present invention is directed to a system in which there are a plurality of controlled variables of a controlled object, a plurality of manipulated variables for controlling these, and The present invention relates to an apparatus for determining optimal control constants for a multi-input/output control system in which changes in quantities each affect a plurality of control quantities.

(Prior Art) Conventionally, in this type of control device, a manipulated variable and a controlled variable whose control effect mainly appears are considered as sets, and a control method of one person's output is applied to each set. However, in this type of control device, the control system is designed by focusing only on one manipulated variable that is paired with one controlled variable and ignoring the influence of other manipulated variables. In some cases, the control performance became worse, or in extreme cases, it became unstable.

On the other hand, a control method based on a state space method is known as a control method when the controlled object is a multi-input/output system. This control method consists of a controlled object 1 and a controller 2 as shown in Figure 2.
It is expressed in the form of In other words, this control method uses the state variable x (
t) is multiplied by a constant matrix by 1 and fed back to improve the characteristics.Since the design takes into account the internal characteristics of the controlled object 1, a stable control system can be designed even if it is a multi-input/output system. can do. Furthermore, since the controller 2 has a structure in which the integral operation works as long as there is a difference between the target value r (t) and the control fay(t), a control system without offset can be constructed.

Regarding this control method, for example, "General method for linear multi-input/output optimal tracking control system" (Proceedings of the Society of Instrument and Control Engineers, Vol. 1)
4, No. 4, August 1973).

In this method, the following scalar control index J is set from a plurality of control variables and manipulated variables, and control is performed so that this becomes the minimum value.

Here, v (t) = u (t), and the variables calculated based on the values in the equilibrium state are expressed with the subscript e. That is, yθ (1) is the control deviation vector. Furthermore, the Q and H matrices are control coordination weights for each controlled variable and each manipulated variable.

In this case, the weight matrices Q and R are design parameters, and Q is set so as to satisfy the control system design specifications expressed by rise time, decay rate, establishment time, etc.

R must be adjusted. If the Q and H matrices are chosen as fixed symmetric matrices, it is possible to design a stable control system, but it is not clear how to express the above design specifications as an ffi-cut matrix. . For this reason, we had to make adjustments through trial and error while performing time response simulations.

(Problem to be solved by the invention) As described above, in the conventional method, the design parameter Q
, and how to set R has not been clarified, so if the time response simulation of the control system does not satisfy the design specifications, Q.
, it is necessary to repeatedly adjust the values of R by trial and error, and there is a problem in that it takes a great deal of time and effort to set the control constants of the control system.

It is an object of the present invention to provide a control constant determining device capable of adjusting the values of Q and H without performing time response simulation and efficiently determining the optimum control constants of a control system. purpose.

[Structure of the Invention] (Means for Solving the Problems) A control constant determining device according to the present invention includes a dynamic characteristic input section of a controlled object, a setting section for weight matrices Q and H, a Riccati equation solving section, Control constant calculation section, closed loop control system pole/zero calculation section, pole/zero values, Q.

A reasoning section that infers whether the specifications can be satisfied from the value of R, design specifications, etc., and if it is not satisfied, how to adjust the values of Q and R, and the necessary and sufficient parts for that inference. It has a database that stores knowledge information.

(effect) In the device according to the invention, a dynamic characteristic input part and a Q.

Based on the information from the R setting section, the Riccati equation solving section solves the equation, and the results are manually input to the control constant calculation section. The control constant calculation section performs certain four arithmetic operations, and (
1) Find the control constant that minimizes the control index in the equation. Based on this result, the closed-loop control system pole/zero calculation section calculates the poles and zeros of the closed-loop control system, and inputs the results to the inference section. The inference section inputs the poles and zeros of the configured closed-loop control system and the design specifications output from the control design specification setting section, and infers whether or not the specifications are satisfied, and if not, calculates the weight matrices Q and H. Inference is made as to how large or small the value of each element should be adjusted.

This inference section uses a database that stores knowledge information necessary and sufficient for inference according to the purpose of inference.

Due to the above, time simulation is not possible.Q.

Compared to the conventional method that searches for the value of R, the time required to determine the optimal control constant can be shortened, and
, H can be adjusted automatically.

(Example) Hereinafter, the present invention will be explained in detail according to examples.

FIG. 1 is a block diagram showing the functional configuration of a control constant windowing device according to an embodiment of the present invention. In the figure, 3 is a dynamic characteristic input section that inputs the dynamic characteristics of the controlled object, and 4 is a weight matrix Q.
, R setting section 5 creates the Riccati equation necessary to calculate the control constant that minimizes the control index of equation (1) from the data of these input section 3 and setting section 4, An equation solving section 6 solves this, a control constant computing section 6 executes predetermined matrix arithmetic 7a using the solution found in the solution finding section 5 to determine control constants; a closed-loop control system pole/zero calculation unit that calculates the poles and zeros of the closed-loop control system configured using the control constants;
8 inputs the values of poles and zeros obtained by the calculation unit 7 and the specifications set by the control system design specification setting unit 10, and infers whether the specifications are met or not. If not, further processing is performed. This is an inference unit that estimates how much the values of each element of the weight matrices Q and H should be adjusted in either direction, large or small, and 9 is a database that stores knowledge information necessary and sufficient for the inference.

In such a configuration, the dynamic characteristic input unit 3 manually inputs the dynamic characteristic of the controlled object using the following state space representation.

x (t) = A X (t) 1 B u (t)
=12)Y (t) -Cx (t) +D
u (t) = 13) Here, the controlled object has m inputs and r outputs (m and r are positive integers, m≧r≧1), and the dimension of the state ff1x(t) is n.

The weight matrices Q and R setting section 4 initially set unit matrices for both. In the Riccati equation solving section 5, A, B, C,
Read D, Q, and R and solve the following Norikatti equation.

The method for solving equation (4) is described in detail in, for example, ``Matrix Theory for System Control'' (Kodama, Suda; Society of Instrument and Control Engineers). The control law that minimizes the control index equation (1) is given by the following equation.

...(5) LL″″Klv11+ to 2v21 °°(6
) L 2-K IV 21 + K 2 V 22
...(7)...(8) The control constant calculation unit 6 outputs the state feedback coefficient L1 and the integral coefficient L2. The closed-loop system pole/zero calculation unit 7 uses the control constant L1. determined in the previous process. The poles of the closed loop control system constructed by the control law of equation (5) using L2 are
Calculate zero. The values of the poles and zeros of the closed loop control system determined by the calculation section 7 are input to the inference section 8.

The inference unit 8 infers dynamic characteristics such as coordination, damping performance, stability, and oscillation of the closed-loop control system from these positions of poles and zeros on the complex plane. An example of the evaluation criteria is given below. Assume that the pole relatively close to the imaginary axis (this is called the main pole) is λ1−−α1±j. At this time, α1>β1. αj 〉ε(ε
>o), it is known that the response characteristics are excellent in coordination, damping, etc., and there is virtually no vibration. Here, ε
is a value determined by the control design specifications and the characteristics of the controlled object itself. Furthermore, it is not necessarily necessary that αl>βl. The dynamic characteristics of a closed-loop control system are mostly controlled by the poles, but the amount of overshoot, undershoot, etc. is determined by the position of the zero point. for example,
If there is a zero point on a semiplane on the complex plane, there will be an inverse response during the step response. Even on the left half plane, it is known that the amount of overshoot increases if the pole is closer to the imaginary axis (to the right) than the main pole.

The criteria for dynamic property inference given in the example above are only examples with respect to the main poles. Therefore, in this embodiment, while referring to the database 9 that stores knowledge information such as the relationship between the positions of poles and zeros and the dynamic characteristics of the closed loop control system,
The dynamic characteristics of the constructed control system are inferred without time response simulation. As a result of this inference, if the specifications are not satisfied, the inference unit 8 further infers adjustment of the weight matrices Q and R in the control index of equation (1). Knowledge information necessary and sufficient for this inference is systematized in the form of rules, for example, and stored in the database 9. In this process, the weight matrix Q,
R will be adjusted to meet design specifications.

The effects of this embodiment will be specifically explained below.

The controlled object is a generator that controls two controlled variables using two manipulated variables, and is a multi-input/output system that is internally interfering with each other. This controlled object has strong oscillatory properties, and conventionally it took a long time to determine the weight matrices Q and H.

FIG. 3 shows a response waveform of a controlled variable when one manipulated variable is changed stepwise. The response waveform 11 indicates that the two controlled variables are interfering with each other, and at the same time it is an oscillatory controlled object.
It can be seen from a and 12a.

Therefore, as a control system design specification, we designed a non-interfering control system that suppresses vibration and has excellent responsiveness. Without running a time response simulation, the dynamic characteristics were inferred from the positions of the poles and zeros of the closed-loop control system, and Q and R were adjusted based on rules and experience.

FIG. 4 shows the target value step response waveform of the closed loop control system based on the optimal adjustment value thus obtained.

At time zero, control ff111b quickly follows, and
It can be seen that the other control amount 12b is hardly disturbed. Moreover, the step response does not oscillate as specified. To obtain such a control system, the adjustment of Q and R only required several iterations.

[Effect of the invention] As described above, according to the present invention, the weight matrix Q.

Since the adjustment of R is performed by inferring whether specifications are met from the positions of poles and zeros in the closed-loop control system, it is no longer necessary to perform time response simulations under various conditions as in the past. Moreover, Q.

In R's weight matrix adjustment, inferences are made while referring to a database that stores necessary and sufficient knowledge information, so it is possible to automate the adjustment by making use of the know-how of experts. Therefore, by using the control constant determination device according to the present invention, it is possible to significantly reduce time and costs.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a control constant determining device according to an embodiment of the present invention, FIG. 2 is a diagram showing an example of the configuration of a follow-up control system, and FIG. 3 is a block diagram showing a configuration example of a follow-up control system. FIG. 3 is a block diagram showing a configuration example of a follow-up control system. Fig. 4 is a diagram showing an example of the response of the controlled variable when the target value of the control system designed using the device of the embodiment is changed in a stepwise manner. . DESCRIPTION OF SYMBOLS 1... Controlled object, 2... Controller, 3... Dynamic characteristic input section, 4... Weight matrix setting section, 5... Riccati equation solving section, 6... Control constant calculation section, 7... Closed loop control system pole/zero calculation unit, 8... Inference unit, 9... Database, 10... Control system design specification setting unit. Applicant's agent Patent attorney Takehiko Suzue TIME Figure 3 TIME Figure 4

Claims (1)

[Claims] A dynamic characteristic input section of the controlled object, a weight matrix setting section, a solution section that receives data from these input sections and the setting section, constructs a Riccati equation, and calculates the solution, and uses the solution obtained by this solution section. a control constant calculation unit that calculates control constants that minimize the control index, and a closed-loop control system pole and zero that calculates the poles and zeros of a closed-loop control system configured using the control constants obtained by this calculation unit. The calculation unit inputs the values of poles and zeros of the closed-loop control system obtained by this calculation unit and the design specifications set in the control system design specification setting unit, and the specifications are created based on the knowledge information stored in the database. 1. A control constant determination device for a control system, comprising: an inference unit that infers whether or not the condition is satisfied, and infers how to adjust the value of a weight matrix when the condition is not satisfied.