JPH0540505A - Fuzzy inference control method - Google Patents

Abstract

PURPOSE:To reduce overshoot without sacrificing the speed of the rise of PID control in the past and in addition, through simple calculation. CONSTITUTION:The deviation of the value PV of a controlled system at present from a set value SP is inputted to a PID arithmetic part 1 and a fuzzy inference part 4. Then, an interval between a control starting point and the set value is divided into three control regions, and two religions near the control starting point and the set value are controlled by PID arithmetic operation, and simultaneously, the region positioned at a middle part is controlled by fuzzy inference that a controlled variable calculated by the arithmetic operation in the first PID control is inputted as a variable, and fuzzy inference output is obtained by multiplying this input variable by prescribed gain. Since the rise is expedited at the control starting point, and the true controlled variable is set by the fuzzy inference in the middle region, and further, the controlled system reaches the set value by using again the PID control in the last region, the overshoot is prevented.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a control system, particularly a method of controlling temperature, flow rate or position by fuzzy reasoning, and can be applied to control of temperature, flow rate or position.

[0002]

2. Description of the Related Art Conventionally, overshoot is often a problem especially in control systems. In order to suppress this overshoot, trial-and-error has been performed by adjusting the parameter of the PID constant in the control based on experience and intuition, especially in the PID control. That is, in this PID control, if the PID constant is set with an emphasis on the rise of control,
As shown in FIG. 1, overshoot is likely to occur, and conversely, if the PID constant is set so as to suppress overshoot, it takes a long time to reach the set value as shown in FIG.

[0003]

The problem to be solved by the present invention is that the correction of the PID constant based on the intuition is extremely complicated and requires specialized knowledge.

[0004]

A fuzzy inference control method according to the present invention divides a control start point and a set value into three control areas, and performs PID calculation on two areas close to the control start point and the set value. Control by the fuzzy inference that inputs the control amount calculated by the calculation in the first PID control as a variable and multiplies this input amount by a predetermined gain to obtain the fuzzy inference output. Is.

[0005]

When the control is started, the rise is accelerated by the PID control, and in the intermediate region, the change amount is set by multiplying the control amount of the PID control by a predetermined count by fuzzy inference, and this is set as the true control amount. Further, in the final region, the set value is again reached by the PID control to prevent overshoot.

[0006]

The fuzzy inference control method of the present invention will be described below. That is, in the configuration diagram shown in FIG. 3, 1 is a PID calculation unit, 2 is a gain setting unit, 3 is a control target, and 4 is a fuzzy inference unit.

The algorithm in this block diagram is as follows. Step 1 First, the deviation x1 = SP-PV of the current value PV of the control target with respect to the set value SP is calculated as PID calculation unit 1 and fuzzy inference unit
To enter. Step 2 The PID calculator 1 outputs the first controlled variable MV1. Step 3 The fuzzy inference unit 4 outputs the gain K and inputs it to the gain setting unit 2. Step 4 In the gain setting unit 2, the first control amount MV1 is multiplied by the gain K to obtain the second control amount MV2 = MV1 * K. Step 5 The controlled object 3 is controlled using this as a final control amount.

The fuzzy inference unit 4 is composed of a deviation operation unit 41 and a fuzzy inference operation body 42 as shown in FIG.

The algorithm in this configuration will be described. Step 1 Input deviation x1 = SP-PV to deviation calculator 41. Step 2 This deviation calculator calculates a deviation x2 for fuzzy reasoning from the difference between the set value SP and the control amount PV, normalizes it to 0 to 1 and outputs it. Step 3 This normalized deviation x2 is the fuzzy inference operation body 4
Entered in 2.

Step 4 The fuzzy inference main body 42 infers according to the following rules. First, set the following control rules. R1: If x2 is A1 then y1 = f1 (x2) ... (1) R2: If x2 is A2 then y2 = f2 (x2) ... (2) R3: If x2 is A3 then y3 = f3 (x2) ) ... (3) where f1 is y2 = ax2 + b (where a and b are constants) ... (4) f2 is y1 = 1 ... (5) f3 is y3 = 1 ... (6) A1: x2 is small, A2: x2 is medium, and A3: x2 is large. The values of the constants a and b are values that change depending on the control target, and are determined by trial and error.

The suitability w1, w2, w3 of the antecedent part with respect to the input x2 is w1 = A1 (x2) ... (7) w2 = A2 (x2) ... (8) w3 = A3 (x2) ... -(9) is calculated. The inference result of each rule is directly x
A value obtained by substituting 2 is y1 = f1 (x2) ... (10) y2 = f2 (x2) ... (11) y3 = f3 (x3) ... (12). The overall inference result is the weighted average value. That is, Y = (w1y1 + w2y2 + w3y3) / (w1 + w2 + w3) (12) is calculated, and Y is set as K in FIG. In particular, the arithmetic expression of the consequent part is expressed by a linear expression as shown by f2 (x) = ax + b and is controlled by a simple arithmetic operation.

Assuming that the controlled variable PV is at the point P in FIG. 5, the deviation x1 at this time is x1 = SP-PV. Then, the deviation x1 is normalized to be x2. At this time, in the membership function shown in FIG.
The deviation is small, and the deviation increases as the regions A2 and A3 progress. A1 likeness is 0.7 A2 likeness is 0.3 A3 likeness is 0.0. This becomes an output by fuzzy inference, and is set as the gain K in FIG. Therefore, the final control output M
V2 becomes MV2 = MV1 * K

An example in which a rule is actually created and executed is shown below. The set system is transfer function G (s) = e ↑ −LS /
(1 + T1S) (1 + T2S) where T1 = 6 and T2 =
10, L = 3. Considering application to discrete time control, this was controlled by the fuzzy inference control method in the present invention with an update cycle of 4 seconds.

Correspondence between system states and fuzzy variables is shown in FIGS. 7 and 8. That is, in FIG. 7, the deviation x is 0 <x <
When E1, the applicable area A1 and the deviation x are E1 <x
<E2, the applicable area A2 is set, and the deviation x is E2.
When <x <SP, the applicable area A3 is set. The membership function at this time is as shown in FIG. The rules in FIGS. 7 and 8 are as follows. IF E2 <x THEN Y1 = 1 IF E1 <x <E2 THEN Y2 = ax + b IF E1> x THEN Y1 = 1 Here, E1 and E2 are fuzzy variable application area movement variables, which are changed depending on the system.

As a result, as shown by the symbol (a) in FIG. 9, overshoot occurs only by the PID control that emphasizes rising, but by adding fuzzy control, as shown by the symbol (b) in FIG. Overshoot is suppressed.

[0016]

As described above, the fuzzy inference control method according to the present invention divides the control start point and the set value into three control areas, and performs PID calculation on the two areas close to the control start point and the set value. In addition to controlling by the fuzzy inference, the region located in the middle is input as a variable with the control amount calculated by the calculation in the first PID control and the input amount is multiplied by a predetermined gain to obtain a fuzzy inference output. It is possible to easily remove the overshoot, which was difficult to achieve by the conventional PID control alone. Particularly in the middle region, a high degree of control can be realized with a simple formula, and therefore the amount of calculation can be reduced, which has the advantage that the calculation time can be shortened. Further, since the beginning and the end of the control are controlled by the PID control, the experience of the conventional PID control can be utilized as it is without sacrificing the rising speed, and there is an advantage that the operation is stable.

[Brief description of drawings]

FIG. 1 is a characteristic diagram in PID control that emphasizes rising.

FIG. 2 is a characteristic diagram in PID control that emphasizes overshoot.

FIG. 3 is a configuration diagram showing an embodiment of a fuzzy inference control method of the present invention.

FIG. 4 is a configuration diagram of a fuzzy inference unit in FIG.

FIG. 5 is an operation explanatory view explaining an operation in the fuzzy inference control method of the present invention.

FIG. 6 is a membership function diagram in the fuzzy inference control method of the present invention.

FIG. 7 is a control area diagram showing a relationship between a system state and a control application area in the fuzzy inference control method of the present invention.

FIG. 8 is a membership function diagram corresponding to FIG.

FIG. 9 is a comparative characteristic diagram showing the relationship between the characteristic in the conventional control and the characteristic in the fuzzy inference control method of the present invention.

[Explanation of symbols]

 1 PID calculation unit 2 Gain setting unit 3 Control target 4 Fuzzy inference unit 41 Fuzzy inference unit Deviation calculation unit 42 Fuzzy inference calculation body

Claims (1)

[Claims] 1. A first control region having a relatively large deviation between the control start point and the set value, which is close to the set point by a predetermined value from the control start point, and a predetermined range from the first control region. Is divided into a second control area having an intermediate deviation closer to the set point by a value of and a third control area having a relatively smaller deviation from the second control area to the set point, and the first area Is controlled by a PID calculation, and in the second region, the control amount calculated by the PID calculation in the first region is input as a variable, and the input amount is multiplied by a predetermined gain to obtain a fuzzy inference output. Further, in the third control area, the fuzzy reasoning control method is characterized in that control is performed again by PID calculation based on the second output.