JPH06110503A - Device and method for evaluating control characteristic and device and method for optimizing control characteristic - Google Patents

Abstract

PURPOSE:To enable evaluation corresponding to the purpose of control by respectively weighting and evaluating the initial part and convergent part of a controlled variable separated by a separation algorithm. CONSTITUTION:A control evaluation part is provided with an SP/PV input part 30 for inputting a target value SP and a process value PV and a calculation part 31 for calculating control deviation (e) used for the separation algorithm or differentiated values e', e'' and e''' of this deviation. A separation judge part 32 judges the separate point of control evaluation corresponding to the separate algorithm. An initial part control evaluation part 33 performs the control evaluation of the initial part separated by the separate algorithm, and a convergent part control evaluation part 34 performs the control evaluation of the convergent part. Then, a control evaluation synthesizing part 35 synthesizes an entire control evaluation value from the control evaluation of the initial part and convergent part. Then, a control evaluation output part 36 outputs the provided control evaluation value. Thus, control regarding start as important or control regarding target value setting can be selectively evaluated.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a control characteristic evaluation device and a control characteristic optimization device for evaluating and optimizing control characteristics in regulatory control such as process control.

[0002]

2. Description of the Related Art Generally, as an evaluation function for evaluating control characteristics in process control or the like, a square integral of deviation or an absolute value integral (deviation area) is often used over the entire range from control to settling of a target value. To be In such an evaluation function, a discontinuous surface does not occur in the evaluation function space spanned by the tuning parameters, and an optimum value search can be easily performed by an iterative method such as a simplex method.

[0003]

However, in such an evaluation function, the entire range of control is uniformly emphasized, and for example, control such that importance is attached to rising of control or conversely, importance is attached to settling of target value. Can not be evaluated.
Therefore, there is a problem that it is not always possible to optimize the parameters that match the control purpose. That is, for example, when the parameters (Kp, Ti, Td) of the PID control are optimized by these evaluation functions, it is possible to optimize only the control including overshoot of several percent.
However, I hate overshoot, but I want to speed up the rise while suppressing this to the first order, or conversely, I want to perform tuning that emphasizes the rise even at the expense of settling and overshoot. Such tuning cannot be handled by a mere square integral of the deviation or an evaluation function based on the deviation area.

[0004]

In order to solve such a problem, the present invention provides a separation algorithm for separating the change of the control amount at the time of changing the target value of the regulatory control into an initial motion portion and a convergence portion based on an index. Is provided, and an evaluation function unit that stores an evaluation function for weighting and evaluating the initial motion portion and the converged portion separated by the separation algorithm is provided. In addition, the change of the control amount when the target value of the regulatory control is changed is separated into the initial motion part and the convergence part based on the index, and the separated initial motion part and the convergence part are each weighted and evaluated. This is a method for evaluating control characteristics. Also, a separation algorithm unit that stores a separation algorithm that separates the change in the control amount when the target value of the regulatory control is changed into an initial motion part and a convergence part based on an index, and an initial motion part and a convergence part that are separated by the separation algorithm. An evaluation function unit for storing an evaluation function for weighting and evaluation, and a means for optimizing a parameter for determining a control characteristic by the simplex method using the separation algorithm and the evaluation function are provided. In addition, the change of the control amount when the target value of the regulatory control is changed is separated into the initial motion part and the convergence part based on the index, and the separated initial motion part and the convergence part are each weighted and evaluated. , Is a method of optimizing the parameters that determine the control characteristics by the simplex method.

[0005]

The initial moving part and the converging part of the control amount change separated by the separation algorithm are weighted and evaluated. As a result, when performing target value tracking control, for example, it becomes possible to selectively perform control evaluation that emphasizes rising of control or control that emphasizes target value settling, and performs evaluation that matches the control purpose. be able to. The parameters that determine the control characteristics are optimized by the simplex method using the separation algorithm and the evaluation function. As a result, optimum control can be performed according to the purpose.

[0006]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the present invention.
The apparatus of this embodiment is used, for example, in a process control apparatus to perform parameter tuning so that optimum controller parameters can be obtained. In FIG. 1, 1 is a control calculation unit, 2 is a control parameter calculation unit, 3 is a control evaluation unit, and 4 is a control target.

Here, the control calculation unit 1 inputs a target value SP, and the input target value SP is used as a control parameter from the control parameter calculation unit 2 and a control amount fed back from the controlled object 4, that is, a process value PV. Based on the calculated value, the operation amount MV is output to the controlled object 4. Also,
The control parameter calculation unit 2 calculates the optimum control parameter based on the control evaluation value J from the control evaluation unit 3 and the simplex method described later, and outputs it to the control calculation unit 1. In addition,
The control evaluation unit 3 inputs the target value SP and the process value PV from the controlled object 4, and evaluates these values for each of an initial motion portion and a convergence portion, which are separated according to a separation algorithm, which will be described later, and a control evaluation value. It is output as J to the control parameter calculation unit 2.

Generally, when the target value SP changes, the manipulated variable M is set so that the process value PV follows the change in the target value SP.
V is changed. As a result, the process value PV changes. In such a case, in this case, the rising portion (that is, the initial movement portion) where the process value PV starts to change toward the target value SP and the process value PV are the target. Value SP
The control is evaluated by dividing it into a settling portion (that is, a convergent portion).

Assuming that the deviation between the target value SP and the process value PV is e, "d ² e / dt ² = 0 and d ³ e
"The time when the condition of / dt ³ > 0 is first met" is defined as the separation point between the initial motion portion and the convergence portion, and the reason for this definition will be described below with reference to FIG. That is, in FIG. 4A, when the dotted line part is the separation line of the initial motion part and the convergence part, the control operation of the initial motion part from the control rising time to the separation line is the operation amount according to the change of the target value SP. Control is performed so that the increase amount of MV is large. As a result, the process value PV quickly approaches the target value SP. When the process value PV approaches the target value SP, the control value MV is controlled so as to be decreased so that the process value PV does not exceed the target value SP. On the other hand, in the control operation of the converged portion after the separation line, the increase / decrease amount of the manipulated variable MV is finely controlled so that the process value PV is settled to the target value SP. In the actual process, since there is a delay time between the change in the manipulated variable MV and the change in the controlled variable PV, the manipulated variable MV has entered the convergence operation before the separation point. Here, the separation point indicates the point of distinction between the initial movement portion and the convergence portion as seen from the change in the control amount PV that occurs as a result of the control operation (that is, the change in the operation amount MV).

That is, in FIG. 4A, the fluctuation of the process value PV is large at the start of the control in the initial motion portion, and the fluctuation is small as the process value PV approaches the target value SP (in this case, the value "4"). Become. After that, in the converging portion, the process value PV changes in a vibrationally small manner. Therefore, in order to find the separation point, the process value PV is first-order differentiated (dPV / dt) with respect to time. Then, a curve as shown in FIG. 4B is obtained. Then, when second differentiation is further performed (d ² PV / dt ² ), a curve as shown in FIG. 4C is obtained. Here, in the curve shown in FIG.
The points a, b, and c are d ² PV / dt which are candidates for separation points.
^A point that satisfies ² = 0, that is, an inflection point. Of these inflection points, the only inflection point that satisfies the condition that the curve falls and rises again, that is, the inflection point that satisfies the condition of d ³ PV / dt ³ > 0 for the first time after the start of control is b point, Therefore, this point b is a separation point that separates the initial movement portion and the convergence portion. Such a separation point b is the “time when the conditions of d ² e / dt ² = 0 and d ³ e / dt ³ > 0 are first met” defined above.

As described above, the control is evaluated by separating the initial moving portion and the convergent portion at the separation point b. In this case, the evaluation function J is expressed by the following equation. That is, J = k * S ₁ + (1-k) * S ₂ (1) (where 0 <k <1)

Here, S ₁ is the absolute value integral value of the deviation of the initial motion portion, S ₂ is the absolute value integral value of the deviation of the convergent portion, and k is the tuning coefficient. This tuning coefficient k is a coefficient that determines the form of control, and if k is set to a value close to "1", control with emphasis on the initial motion portion is performed, and k is "0".
If the value is close to, the control focusing on the converged portion is performed. Therefore, it becomes possible to specify a control shape that could not be evaluated by the conventional control evaluation based on absolute value integration (that is, deviation area) over the entire control range. When k = 0.5, it is equivalent to the conventional control evaluation based on the deviation area.

FIG. 2 is a block diagram of the control evaluation unit 3 capable of designating such a control shape, and FIG. 3 is a flow chart showing its operation. That is, in FIG. 2, the control evaluation unit 3 inputs the target value SP and the process value PV to SP,
PV input unit (hereinafter referred to as input unit) 30, control deviation e used in the separation algorithm and differential values e ′, e ″ of this deviation,
e, e ′, e ″, e ′ ″ calculation unit (hereinafter, calculation unit) 31 for calculating e ″ ′, separation determination unit 32 for determining a separation point of control evaluation by a separation algorithm, separation by a separation algorithm Initial control part control evaluation unit 33 that performs control evaluation of the initial motion part, convergent part control evaluation unit 34 that performs control evaluation of the convergence part separated by the separation algorithm, overall control evaluation from control evaluation of the initial motion part and the convergence part It is composed of a control evaluation combining unit 35 that combines values and a control evaluation output unit 36 that outputs the obtained control evaluation value.

Here, a specific control operation of the control evaluation unit 3 will be described with reference to the flowchart of FIG. That is, it is detected in step 100 whether or not the value of the target value SP is changed, and in step 101, the absolute value integral values (ie, deviation area values) S ₁ and S ₂ of the deviations of the initial motion portion and the convergence portion are respectively detected. Are set to “0” respectively. After that, the input unit 30 inputs each value of the target value SP and the process value PV (step 102), and the calculation unit 31 deviates from the input target value SP and process value PV the deviation e and the differential value e ′ of this deviation. , E ″, e ′ ″ are determined (step 103), and the separation determination unit 32 determines whether the portion currently controlled is the initial movement portion or the convergence portion based on the differential value of these deviations (step 103). 104).

That is, if the current control is before the passage of the separation point b, it means that the control of the initial motion portion is being performed, so that the deviation area S of the initial motion portion in the initial motion portion control evaluation unit 33. ₁ is calculated (step 105).
Then, in step 107, has the control reached the convergence point at which the process value PV is settled to the target value SP (that is, the time when the absolute value of the control deviation e does not exceed the settling range m during a certain period t)? In this case, since the separation point b has not been reached yet, the process returns to step 102 and the target value SP and the process value PV are input again. In this way, the target value SP and the process value PV are input until the process control reaches the separation point b, and the deviation area S ₁ of the initial motion portion is calculated based on these values.

The target value SP and the process value PV are input, and it is determined that the process control currently being performed has passed the separation point b from the result of the calculation of the deviation or differential value thereof (step 1).
04), the deviation area S ₂ of the convergent portion is calculated in the convergent portion control evaluation unit 34 (step 106). And
The input of the target value SP and the process value PV in this convergence portion and the calculation of the deviation area S ₂ thereof are continued until the control reaches the convergence point. After that, when the passage of the convergence point is determined in step 107, the control evaluation synthesis unit 35 uses the evaluation function J as shown in the equation (1) to determine the deviation area S ₁ of the initial portion and the deviation area S _{2 of the} convergence portion. The overall control evaluation value is calculated from (step 108), and this control evaluation value is output from the control evaluation output unit 36 to the control parameter calculation unit 2 (step 109). in this way,
According to the present invention, when the process control is evaluated, the initial part and the convergent part are separately evaluated.

Next, FIG. 5 is a diagram showing the concept of the control evaluation function curved surface. That is, here, with respect to the second-order lag system with waste time, that is, when the transfer function is Gp, Gp = 4.0exp (−Ls) / (12.8s + 1) (20.0 + 1) (2) The following PI control is performed. ΔMV = Kp (Δe + 1 / T · e) (however, e = SP-PV) (3)

When a set of parameters Kp, Ti is determined here, the step response by PI control for one process is uniquely determined, and therefore the control evaluation value by the control evaluation function is also uniquely determined. That is, a curved surface based on the control evaluation function can be assumed for the two-variable space determined by the parameters Kp and Ti. Next, a method of optimizing the parameters Kp and Ti by the simplex method will be described.
In the simplex method, a plurality of points (points A, B, and C in FIG. 5) are considered with respect to the two-variable space of parameters Kp and Ti. Then, among these points, the centroid point composed of the point with the largest evaluation value, that is, the point with the poor evaluation value (point C) and the other points (points A and B) is calculated, and these two points are externally divided. To move to the point (C 'point)
Change i. By sequentially moving the worst evaluation value in this way, the set of these points is converged on the minimum point of the control evaluation function, and as a result, the optimum point of the control evaluation value, that is, the optimum parameter Kp, Ti. Will be obtained.

In the conventional evaluation function, the curved surface defined by the control evaluation function is unique, and therefore the optimum control point cannot be moved. However, in the present invention, the value of the tuning coefficient k in the evaluation function can be changed. The curved surface can be deformed by the control evaluation function, and thus the optimum point can be moved.

FIG. 6 is a diagram showing an example of a step response when the gain Kp of PI control and the integration time Ti are optimized while changing the tuning coefficient k in the evaluation function J. Here, when the tuning coefficient k is reduced, the process value PV gently rises but quickly approaches the target value SP, as shown in FIGS. 6A and 6B, since the focus is on the convergent portion. On the other hand, when the tuning coefficient k is increased, the initial movement part is emphasized, and
As shown in (e), the process value PV rapidly rises and hunts slightly to approach the target value SP.

[0021]

As described above, according to the present invention,
Since the initial part and the convergent part of the control amount separated by the separation algorithm are weighted and evaluated,
When performing target value tracking control, for example, it becomes possible to selectively evaluate the control that emphasizes the start-up of the control or the control that emphasizes the target value settling. Therefore, perform the evaluation that matches the control purpose. You can In addition, since the parameters that determine the control characteristics are optimized by the simplex method using the above separation algorithm and evaluation function,
Optimal control can be performed according to the purpose.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a detailed block diagram of a control evaluation unit in the apparatus of the embodiment.

FIG. 3 is a flowchart showing an operation of the control evaluation unit.

FIG. 4 is a diagram illustrating a separation point between an initial motion portion and a convergence portion in process control.

FIG. 5 is a diagram showing a concept of a control evaluation curved surface by a control parameter space and parameter optimization by a simplex method.

FIG. 6 is a graph showing changes in the step response of the PI control optimized by the evaluation parameter k of the apparatus of the above embodiment.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 Control calculation part 2 Control parameter calculation part 3 Control evaluation part 4 Control object 30 SP, PV input part (input part) 31 e, e ', e'',e''' calculation part (calculation part) 32 Separation determination part 33 initial motion partial control evaluation unit 34 convergent partial control evaluation unit 35 control evaluation synthesis unit 36 control evaluation output unit SP target value MV manipulated variable PV process value J control evaluation value

Claims (4)

[Claims] 1. A separation algorithm unit that stores a separation algorithm that separates a change in the control amount when the target value of the regulatory control is changed into an initial motion portion and a convergence portion based on an index, and an initial motion that is separated by the separation algorithm. A control characteristic evaluation device comprising: an evaluation function unit that stores an evaluation function for weighting and evaluating a part and a converging part, and the control characteristic is evaluated by the evaluation function. 2. The change of the control amount when the target value of the regulatory control is changed is separated into an initial motion part and a convergence part based on an index, and the separated initial motion part and the convergence part are respectively weighted. And a control characteristic evaluation method for evaluating control characteristics. 3. A separation algorithm unit that stores a separation algorithm that separates a change in the control amount when the target value of the regulatory control is changed into an initial motion portion and a convergence portion based on an index, and an initial motion that is separated by the separation algorithm. An evaluation function unit for storing an evaluation function for weighting and evaluating the part and the convergent part, and means for optimizing parameters for determining control characteristics by the simplex method using the separation algorithm and the evaluation function are provided. A control characteristic optimizing device characterized by the above. 4. The change of the control amount when the target value of the regulatory control is changed is separated into an initial motion part and a convergence part based on an index, and the separated initial motion part and the convergence part are weighted respectively. A control characteristic optimization method that evaluates parameters and optimizes the parameters that determine the control characteristics by the simplex method.