JPH04352003A - Fuzzy control device - Google Patents

Abstract

PURPOSE:To obtain a fuzzy control device capable of automating the adjustment of control rules and capable of being applied to a plant in which the same operation is not repeated. CONSTITUTION:The fuzzy control device is provided with preprocessing parts 11, 21 for calculating a control deviation between a target value and a control variable and its one-step difference, a normalizing part 22 for normalizing the control deviation and the one-step difference, an inference part 23 for executing the fuzzy inference operation of the normalized values and converting the obtained value into the normalized value of a change in a manipulated variable, an output part 24 for calculating the change in the manipulated variable based upon the normalized value, an integrating part 14 for integrating the change and calculating the manipulated variable, a signal generating means 18 for generating a test signal for changing the manipulated variable to be controlled by a prescribed value, and a gain control means 19 for calculating a control parameter based upon the manipulated variable changed by the test signal and a control variable corresponding to the manipulated variable and controlling the gain of the normalizing part 22 and the output part 24 by the control parameter.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fuzzy control device that expresses complex and ambiguous control laws using IF-THEN rules and performs control based on the IF-THEN rules.

0002

[Prior art] Complex and ambiguous control laws are processed using IF-T.
A fuzzy control device is known that is expressed in HEN rules and performs fuzzy control based on this control law.

[0003] In this fuzzy control device, a target value and a control amount are input to a preprocessing section, a first-order difference between the control deviation and the control deviation is calculated, and the first-order difference between the calculated control deviation and the control deviation is
The floor differences are further input to the normalization section and normalized there. The value normalized by the normalization unit is input to the inference unit, where fuzzy inference calculation is performed according to a predetermined control law to obtain a normalized value for the change in the manipulated variable. The obtained normalized value is input to the output section to calculate the change in the manipulated variable, and the change is integrated by the integrating section to obtain the manipulated variable. The configuration is such that the manipulated variable obtained in this manner is given to the controlled object.

[0004] The control performance of the fuzzy control device configured as described above greatly depends on fuzzy control parameters such as control rules, membership functions, and input and output gains. However, in most cases, the operator sets the optimal control parameters through trial and error, which requires a lot of effort.

[0005] Several methods have been proposed as fuzzy control devices that improve these. For example, Proceedings of the Society of Instrument and Control Engineers VOL 1.20 NO8 P720-7
26 Or VOL1.24 NO2 P191-
It is 197 mag.

However, in the above-mentioned fuzzy control device, since the control parameters are adjusted by iterative learning,
It is difficult to apply this method to actual plants where the same operation is rarely repeated over and over again.

[0007]

[Problems to be Solved by the Invention] As described above, conventional fuzzy control devices require a lot of effort to set fuzzy control parameters, and cannot be applied to actual plants where the same operation is not repeated. There were some difficult issues.

The present invention has been made in view of the above-mentioned circumstances, and it is an object of the present invention to provide a fuzzy control device that can reduce the labor of the operator and can be applied to plants where the same operation is not repeated.

[0009]

[Means for Solving the Problems] In order to achieve the above object, the present invention provides a preprocessing unit that calculates a control deviation between a target value and a controlled amount and a first-order difference of the control deviation, and A normalization unit normalizes the control deviation and its first difference calculated by using a predetermined control parameter, and the normalized value in this normalization unit is subjected to fuzzy inference calculation according to a predetermined control law to obtain a manipulated variable. an inference unit that calculates the normalized value of the change in the amount of change, an output unit that calculates the change in the manipulated variable using a predetermined control parameter from the normalized value determined by the inference unit, and an operation calculated by the output unit. In the fuzzy control device, the fuzzy control device includes an integrating section that calculates the manipulated variable for the controlled object by integrating changes in the amount, at least for gain adjustment of the normalizing section and the output section. a signal generating means that generates a test signal to be changed by a predetermined amount, a control parameter that calculates each of the control parameters based on a manipulated variable that is changed by the test signal, and a controlled variable that corresponds to the manipulated variable, and the calculated control parameters; The present invention is configured to include the normalization section and gain adjustment means for adjusting the gain of the output section.

0010

According to the present invention, a test signal is input from the signal generating means to the control system of the fuzzy control device and is added to the manipulated variable determined by the integrating section. The controlled object is operated with the operation amount to which this test signal is added, and the corresponding control amount is input to the gain adjustment means and the preprocessing section, respectively. In the gain adjustment means, a control parameter is determined based on the manipulated variable to which the test signal is applied and the controlled variable corresponding to the manipulated variable, and the control parameter is reset to the normalization section and the output section so that both Gain is adjusted. By inputting the test signal in this way, the control parameters to be set in the normalization section and output section are automatically calculated, which reduces the labor of the operator and is applicable to plants that do not operate repeatedly. can.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows functional blocks of a fuzzy control device according to this embodiment.

[0012] This fuzzy control device has a target value r(t)
and the control amount y(t) are given to the subtracter 11, and the deviation e of the control amount with respect to the target value is calculated. The deviation e calculated by the subtracter 11 is converted into a discrete signal via the sampler 12 and input to the control unit 13. The control unit 13 generates a speed-type manipulated variable du from the input discrete signal of the deviation e by fuzzy inference calculation, and outputs it to the integration unit 14. Integral part 1
4 further integrates the input speed-type manipulated variable du, converts it into a position-type manipulated variable, and outputs it to the adder 15. This adder 1
The position type operation output No. 5 is converted into a continuous signal by the holder 16, and then given to the controlled object 17. The control amount corresponding to the operation amount given to the controlled object 17 is calculated by the subtracter 1
1 is fed back.

In this embodiment, the test signal V for gain adjustment outputted from the signal generating section 18 is input to the adder 15 into which the manipulated output is input, and the test signal V and the manipulated output of the integrating section 14 are , and the control amount y(t
) is input to the gain adjustment section 19. The gain adjustment section 19 calculates optimal control parameters, which will be described later, using a known control constant determination method, and outputs a gain adjustment signal to the control section 13 in accordance with the calculation result. FIG. 2 shows the configuration of the control section 13 in detail.

The control unit 13 includes a first-order difference calculation unit 21 that calculates the first-order difference of the deviation e inputted as a discrete signal, and the first-order difference calculated by the first-order difference calculation unit 21 and the deviation e. A normalization unit 22 normalizes the values by a gain according to the set control parameters, and a normalization unit 22 performs a fuzzy inference calculation on the values normalized by the normalization unit 22 to obtain a normalized value for the change in the manipulated variable. An inference unit 23 that calculates
The output unit 24 converts the normalized value obtained in step 3 into a change in the manipulated variable with a gain according to the set control parameter. The control parameters of the normalization section 22 and the output section 24 are adjusted by the gain adjustment section 19. Next, the operation of this embodiment configured as above will be explained.

The target value r(t) given from the outside and the control amount y(t) which is the output of the controlled object 17 are input to the subtracter 11, and the sampler 12 converts the control amount y(t) into a sample signal y( k), and the control deviation e is calculated using the following equation (1).
(k) and the first-order difference de(k) between its control deviation e(k)
Calculate. e(k)=r(k)-y(k) de(k)=e(k)-e(k-1)...(1)

0
[016] The normalization unit 22 calculates the control deviation e(k) and its first
The floor difference de(k) is input, and the normalized value E of the control deviation and the normalized value DE of the first-order difference of the control deviation are calculated using the following equation (2). E=c1 ×e(k), DE=c2 ×de(
k) ...(2) In equation (2), c1, c2
is the control parameter used for gain adjustment.

The inference unit 23 receives the normalized value E of the control deviation and the first-order difference DE of the control deviation, which are the outputs of the normalization unit 22, and performs fuzzy processing using, for example, the control rule shown in the following equation (3). A normalized value DU of the change in the manipulated variable is calculated by inference. IF (E is NB and DE is NB
) THEN (D is NB) IF
(E is NB and DE is NM)
THEN (Dis is NB)
・
・
・IF(E is PB a
nd DE is PB) THEN (D
is NB)

...(3) However, NB, NM, ..., PB are fuzzy variables. The output unit 16 receives the normalized value DU of the change in the manipulated variable calculated by the inference unit 23, and calculates the change du(k) in the manipulated variable using the following equation (4). du(k)=c3×DU
...(4) c in equation (4)
3 is a control parameter used for gain adjustment of the output section 24. The integration unit 14 receives the change du(k) in the manipulated variable calculated by the output unit 24 as input, and calculates the manipulated variable u(k) (sample value) using the following equation (5). u(k)=u(k-1)+du(k)
...(5) u(k-1) in equation (5) is 1
This is the manipulated variable before the sample time.

On the other hand, the gain adjustment section 19 outputs a gain adjustment command signal to the signal generation section 18 during gain adjustment before plant operation, etc., and the signal generation section 18 outputs a gain adjustment command signal to the adder 15.
outputs a test signal V (hereinafter, the test signal V will be referred to as a "test signal d(k)" in correspondence with the manipulated variable u(k)).

The adder 15 receives the manipulated variable u from the integrating section 14.
(k) and the test signal d(k) from the signal generator 18, and the sum signal is sent to the gain adjuster 19 and the holder 1.
Output to 6. The holder 16 converts the input sum signal into a continuous time signal and uses it as the manipulated variable u(k) for the controlled object 17.
Output to.

As a result, the control object 17 outputs a control amount y(k) corresponding to the manipulated variable u(k) to which the gain adjustment operation is added, and the subtracter 11 and gain adjuster 19
are given to each.

Here, the gain adjustment section 19 calculates the control parameters c1, c2, c3 as follows. That is, the manipulated variable u(k) to which the test signal is applied and the controlled variable y(k) taken in from the controlled object via a sampler (not shown)
k). The controlled object G(s) is calculated using the CHR method (Chien Hrones Reswick) of the PID adjustment formula.
k), then G(s)={R/(1+Ts)}e
-Ls...It is expressed by the equation (6). L
is a dead time, T is a time constant, R is a slope, and s is a Laplace operator. In addition, the generally known digital PI control formula is ΔMVn = MVn - MVn-1
=Kp{(en −en−1 )+(τ/TI
)en...(7) τ represents the sampling time, Kp represents the proportional gain, and TI represents the integration time. Then, from the above equation (2) and the above equation (7), c1 = (Kp/TI)・τ, Tτ=(
Kp/c1)・τ...(8)c2=Kp
...(9), and in the PI mode of the above CHR method, the dead time L, time constant T,
Using slope R, Kp=0.7 L/KL, TI
=2.3 L...(10) is defined. (
8), From equations (9) and (10), c2 = 0.7 L
/KL...(11)2.3 L=(Kp/c1)・
τ...(12). Substituting equation (11) into equation (12), 2.3 L=(1/c1)・(0.7 T/KL
)・τ c1 = (0.7/2.3 )・(0.7
T/KL2)・τ=0.213・(T
τ/KL2) …(
13) Here, K=R. Therefore, c1 =0.213(Tτ/RL2)...(1
4) c2 = 0.7 (T/RL)
...(15) As described above, the control parameters c1, c are determined using the CHR method, which is a PID parameter adjustment rule.
2 can be determined.

The control parameter c3 is determined by adding the operation signal for a predetermined operation from the signal generator 18 to the operation amount.
The manipulated variable of the controlled object is changed by ΔMV in steps, and the ratio before and after the change can be determined based on the change in the controlled variable at that time.

In this way, the control parameters c1,
Once c2 and c3 are determined, the gain adjustment section 19
A gain adjustment signal is output from the normalization section 22 and the output section 24, and the control parameters set in each are changed to the calculated corresponding control parameters and the gain is adjusted.

As described above, according to this embodiment, the controlled object 17
A test signal d(k) for gain adjustment is added to the input signal, and the gain of the normalizing section 22 and the gain of the output section 24 are adjusted based on the manipulated variable and the controlled variable at that time to obtain the optimum controlled variable. Therefore, the gain can be adjusted without repeated learning as in conventional methods, and it can be applied to plants that are rarely operated repeatedly. Further, there is no need to make adjustments through trial and error as in the past, and labor can be reduced.

[0025] In the above embodiment, the CHR method is used to determine the control parameters, but the present invention is not limited to such a method.
) as G(s)=(1/Ts)e−Ls…(16)

Approximating with the following formula, in PI mode, the proportional gain Kp is 0.9 T/L, and the integral time TI is 3.3 L.
Z-N (Ziegler Nichols
) can also be used.

[0027]

As described above, according to the present invention, it is possible to provide a fuzzy control system that can reduce the operator's labor required for adjusting control rules and can be applied to plants where the same operation is not repeated.

[Brief explanation of drawings]

FIG. 1 is an overall functional block diagram of a fuzzy control device according to an embodiment of the present invention.

FIG. 2 is a functional block diagram of a control unit of a fuzzy control device according to an embodiment.

[Explanation of symbols]

11... Subtractor, 12... Sampler, 13... Control unit, 14...
Integrating section, 15... Adder, 17... Controlled object, 18... Signal generating section, 19... Gain adjusting section, 22... Normalizing section, 23... Inference section, 24... Output section.

Claims (1)

[Claims] Claim 1: A preprocessing unit that calculates a control deviation between a target value and a controlled amount and a first-order difference of the control deviation; A normalization unit that normalizes using parameters, an inference unit that performs fuzzy inference calculations on the values normalized by this normalization unit according to a predetermined control law to obtain a normalized value for the change in the manipulated variable, and this inference. An output section that calculates the change in the manipulated variable using a predetermined control parameter from the normalized value obtained in the section, and an output section that integrates the change in the manipulated variable calculated by this output section to calculate the manipulated variable for the controlled object. a fuzzy control device comprising: a signal generating means for generating a test signal that changes the manipulated variable of the controlled object by a predetermined amount in order to adjust the gain of at least the normalizing section and the output section; A gain that calculates each of the control parameters based on a manipulated variable changed by a test signal and a controlled variable corresponding to the manipulated variable, and adjusts the gain of the normalization section and the output section using the calculated control parameters. A fuzzy control device comprising: an adjustment means.