JPH04195401A - Fuzzy control system - Google Patents

Abstract

PURPOSE:To evaluate overall judgement results to determine a control instruction by using a second membership function to evaluate a control rule and an internal state transition rule by fuzzy inference operation. CONSTITUTION:This system consists of a fuzzy control executing circuit 1, a fuzzy rule storage circuit 2, a membership value storage circuit 3, a fuzzy internal state storage circuit 4, a control target signal generating circuit 5, and a control object 6 like a plant or a robot. An input value and an internal state are controlled by the control rule related to not only the input value at a settled time but also that at an unsettled time, and the unsettled time is expressed by the second membership function related to time for the purpose of changing the internal state by the state transition rule related to the internal state, and this function is used to evaluate the control rule and the internal state transition rule by fuzzy inference operation. Thus, fuzzy control is realized which is based on rules which evaluate the state of the control object related to unsettled time and elapse of time which a human being has.

Description

[Detailed Description of the Invention] [Objective of the Invention] (Industrial Application Field) The present invention is useful for controlling plants, robots, etc., when the dynamic characteristics of the controlled object are not fully known, or when the dynamic characteristics of the controlled object are The present invention relates to a fuzzy control method, which is a rule-based control method that performs control based on control rules held by a skilled operator, for a controlled object to which it is difficult to apply conventional control theory due to fluctuations in the control system.

(Prior Art) A fuzzy control method is a rule-based control method characterized by expressing control rules held by a skilled operator (expert) using fuzzy theory and determining control commands by fuzzy inference.

This type of control method requires an experienced operator
It is required to be able to sufficiently express the control rules possessed by the system, and to provide appropriate control instructions through inference.

However, the conventional fuzzy control method can only handle the fuzzy internal state at a definite time in which the state of the controlled object at a definite time and the comprehensive decision result of the control at a definite time are stored. Therefore, for example, if the pressure has been high for a while and the control performance has been deteriorating for about 3 minutes, open the control valve a little.
Control instructions are given by evaluating the state of the controlled object and the fuzzy internal state, which is the comprehensive judgment result of the control, regarding uncertain times such as "from a while ago" and "from about 3 minutes ago" in a control rule. This decision cannot be made using conventional fuzzy control methods, which lacks the ability to adequately express the rules of a skilled operator (expert).

(Problems to be Solved by the Invention) As described above, in the conventional fuzzy control method, the state of the controlled object at a definite time and the comprehensive decision result of the control at a definite time are stored at a definite time. In conventional fuzzy control, control instructions can only be determined by evaluating the state of the controlled object at an uncertain time and the fuzzy internal state, which is the comprehensive control decision result. I couldn't do it by method.

The present invention has been made in consideration of the above points, and therefore provides a fuzzy control method that can determine control instructions by evaluating the state of the controlled object and the comprehensive judgment result of control regarding uncertain time and elapsed time. is intended to provide.

[Structure of the Invention] (Means for Solving the Problem) In order to achieve the above-mentioned purpose, the fuzzy control method of the present invention uses a process value measured from a controlled object or a process status value determined by an operator. The control device has an internal state as an input value and stores a determination result regarding the situation of the controlled object, and a first membership regarding the manipulated variable of the controlled object is determined by fuzzy inference calculation according to control rules and state transition rules. A fuzzy control method that calculates a function, determines a manipulated variable from this first membership function to control the controlled object, and changes an internal state, wherein the input value and the internal state are determined. In order to control not only the value at time but also the control rule regarding the human power value at the uncertain time and change the internal state according to the state transition rule regarding the internal state, the uncertain time is set to a second member regarding time. The control rule and the internal state transition rule are expressed by a ship function, and the second membership function is used to evaluate the control rule and the internal state transition rule by fuzzy inference calculation.

(Operation) In the fuzzy control method of the present invention, a fuzzy set (membership function with time as a variable) defined on the time axis is introduced in order to evaluate uncertain times and the passage of time.

Hereinafter, this will be called the fuzzy tense. By describing control rules held by a skilled driver using this fuzzy tense and evaluating them, control based on control rules regarding uncertain times and the passage of time can be realized.

Therefore, it is possible to realize fuzzy control based on rules that humans have for evaluating the state of the controlled object regarding uncertain times and the passage of time.

(Example) Hereinafter, an example of the present invention will be described in detail based on the drawings.

FIG. 1 is a block diagram for explaining the configuration of a control system employing a fuzzy control method according to an embodiment of the present invention.

1 is a fuzzy control execution circuit, 2 is a fuzzy rule storage circuit, 3 is a membership value storage circuit for storing the past certainty of a boozy proposition modified by a fuzzy tense, 4 is a fuzzy internal state storage circuit, 5 Reference numeral 6 indicates a control target signal generation circuit, and 6 indicates a controlled object such as a plant or a robot.

For each control cycle, the fuzzy control execution circuit 1 inputs the control amount signal Y from the controlled object 6 and the target value signal R from the control target signal generation circuit 5, and outputs the manipulated amount signal U by fuzzy control. At this time, the fuzzy control execution circuit 1
All control rules RU and fuzzy internal state transition rules RX stored in the fuzzy control rule storage circuit 2 are read out, and the condition part of the rule is evaluated (if the control deviation E needs to be evaluated in the condition part, E-R-Y
(calculate in advance). For rules that do not include fuzzy tenses, calculations are performed based on inference operations in normal fuzzy control. For rules that include fuzzy tenses, the conditional part of the rule is evaluated using the method described below.

When the controlled object is a system with m inputs and n outputs, the signals U, Y, R, and E at time t are expressed by the following vectors.

Operation amount signal U U (t) = (tz (t) , u; + (t)
, -1u+(t), -1u, (t))" Controlled amount signal Y Y (t)= (yl (t) , "/2 (t)
, ..., yl (t), ..., y, (t))" control target signal R R (t) - (r + (t) , r; + (t)
, -1rl (t), ..., ", (t))" control deviation E E (t) - (el (t) , e, + (t),
−1e+(1,”・,e,(t))” However, el (t)=r+ (t) Yl (
t) Fuzzy internal state variable X (t)= (x+ (t) ,X2 (t)
, -・-1X1 (t), ..., Xk (t)T(
(T represents transposition) As an example of a rule that includes a fuzzy tense, consider the following fuzzy rule.

IP(T(el-A)) and (x,-
B) , then(u, -C) where ASBSC are fuzzy values, each characterized by membership functions fA, fB, fc.

T is a fuzzy tense and is characterized by a membership function f1. At the current time t, the control deviations el (t), Xi (t), (1≦1.) are given, and the certainty of the fuzzy proposition containing the fuzzy tense operator T in the first term of the conditional part of the control rule is calculated. g ((T (e +
−A) ) ) is calculated as follows. Here, t
Let max represent the upper limit of the evaluation range of the fuzzy tense child T (with the current time as the origin 0).

■If the state of the controlled object has been realized at least once,
That is, when T is a fuzzy tense such as "about 5 minutes ago", g ((T (el - A) ) ) -max (m
i n ft 0, tllaX T (t), f ^ (el (
tc-t)))) (1) ■When the state of the controlled object continues, that is, when T is a fuzzy tense such as "from about 5 minutes ago", g ((T (e + −A) ) ) −Σ
(min(frl-0, 1,,.

(t), f^ (el (tc t))))
(2)/Σ ft (t) 1-0, 1. ,.

Equations (1) and (2) show an example, and using fuzzy tenses, (1) and (2) calculation methods for fuzzy tenses other than weapons can also be considered.

In evaluating fuzzy tenses for fuzzy propositions regarding fuzzy internal state variables, it is necessary to match membership functions because fuzzy internal state variables take fuzzy values, that is, membership functions. That is, g = ((T (x r =A)
)), assuming that the value of the fuzzy internal state variable at time t is A'(f^-(x+(t))), the following will be obtained.

■If the fuzzy internal state is realized at least once, that is, when T is a fuzzy tense such as "about 5 minutes ago", g-((T (x+-A))) -max (m
i nl-0, 1,,, x (f T (t), match(f ^ (x+ (
tc -t)), fA-(XI (tc
t))))))'(1)'■If the fuzzy internal state continues, that is, when T is a fuzzy tense such as "from about 5 minutes ago," then g - ((T ( e+ −A) ) )−Σ(min(
fl-0,1,.

7 (t) , mat ch (fA (x, +
(tc-t)), fA・(x+ (tc t))
))))/Σfr (t)1-0.1. ,.

(2)' Here, as the matching calculation matchO, for example, the following calculation can be considered.

matCh (f A (XI), r A't:x
1 )) -max (min (f A
(XI), f A (x 1 ) x 1

(3) matCh (f ^ (XI)
, f ^ (X 1 ))”, l”min
(fA (XI), f ^-(x 1)
) dx.

/, /' fA (x+) dx,
(4) Once the confidence level g of a fuzzy proposition containing a fuzzy tense is calculated, fuzzy logic operations such as logical product and, and inference operations are executed according to the usual fuzzy control method.

Next, the processing procedure will be explained. FIG. 2 is a flowchart of the processing procedure of this embodiment. This process is executed at every sample time.

After starting processing in step SPI at sample time t, the control target value signal R and control amount signal Y are read and stored in step SP2.

If necessary, the control deviation E is calculated from the R and Y stored in step SP3, and the process is repeated twice.

In step SP4, fuzzy rules are read from the fuzzy rule storage circuit 2.

In step SP5, it is determined whether there is a fuzzy proposition regarding a fuzzy internal state variable in the condition part of the rule.

Then, when there is no fuzzy proposition regarding the fuzzy internal state variable, in step SP6, a membership function for evaluating the fuzzy proposition is calculated) and stored.

On the other hand, when there is a fuzzy proposition regarding a fuzzy internal state variable, the value of the fuzzy internal state variable is retrieved from the fuzzy internal state storage unit 4 in step SP7, and the fuzzy proposition is determined using equation (3) or (4). degree (membership value) and store it.

In step SP8, it is determined whether the evaluated fuzzy proposition has a fuzzy tense.

Then, if the fuzzy proposition does not have a fuzzy tense, go to step 5PII.

On the other hand, when the fuzzy proposition has a fuzzy tense, the fuzzy proposition with the fuzzy tense is converted to (1) in step SP9.
Alternatively, the confidence level (membership value) is calculated and stored using equation (2).

Subsequently, in step SP1.0, the current certainty of the fuzzy proposition is stored in the membership value storage circuit 3.

Step SP1. In l, it is determined whether all the fuzzy propositions in the conditional part of the rule have been evaluated, and if YES, in step 5PI2, the certainty factor of the conditional part of the rule is calculated by =-1 and stored.

In step 5P13, it is determined whether the condition parts of all rules have been evaluated, and if YES, step 5P
At step 14, the inference results are integrated using a normal fuzzy inference method, and membership functions are determined and stored.

In step 5P15, it is determined whether the signal is a fuzzy internal state variable or a manipulated variable signal.

In step 5P1B, in the case of a fuzzy internal state variable, the membership function determined in step 5P14 is stored in the fuzzy internal state storage circuit 4.

In step 5P17, in the case of the manipulated variable signal, the manipulated variable signal U is determined and output (based on the normal fuzzy control method).

In step SP1.8, the processing at sample time t is ended and the processing process is placed in a standby state until the next sample time t+i.

An example of a fuzzy tense is shown in Figure 3. For example, "about 5
A control rule such as ``If the pressure is high in the 5 minute term, open the valve a little'' uses the fuzzy tense ``approximately 5 minutes ago term'' in Figure 3 to determine the confidence of the proposition regarding the state of the process ``the pressure is high'' ( The past value of the membership value (membership value) is evaluated using a fuzzy tense to determine the degree of certainty that the conditional part of this rule currently holds true, and a control instruction at the current time is determined.

[Effects of the Invention] The fuzzy control method of the present invention provides a fuzzy control method based on a control rule that evaluates the state of a controlled object in relation to uncertain times and the passage of time that humans have, which cannot be achieved with conventional fuzzy control methods. can be realized. As a result, the control rules possessed by a skilled operator (expert) can be sufficiently expressed, and the scope of application of the fuzzy control method can be expanded.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing the overall configuration of a control system applying the fuzzy control method according to the present invention, Fig. 2 is a flowchart showing the processing procedure of fuzzy control, and Fig. 3 is an example of a membership function representing a fuzzy tense. FIG. DESCRIPTION OF SYMBOLS 1... Fuzzy control execution circuit, 2... Fuzzy rule storage circuit, 3... Membership value storage circuit, 4.
・Fuzzy internal state storage circuit, 5. Control target signal generation circuit, 6. Control target. Applicant: Toshiba Corporation Patent Attorney: Suyama Sa - 1st Membership Value

Claims (1)

[Claims] (1) The process measurement value from the controlled object or the process situation value determined by the operator is used as the input value, and the control device has an internal state in order to store the judgment result regarding the situation of the controlled object, and the control rule and A first membership function regarding the manipulated variable of the controlled object is calculated by a fuzzy inference operation according to a state transition rule, and the manipulated variable is determined from this first membership function to control the controlled object, and the internal state A fuzzy control method that changes the input value and the internal state based on not only the value at a fixed time but also the control rule regarding the input value at an uncertain time, and the state transition regarding the internal state. In order to change the internal state according to a rule, an uncertain time is expressed by a second membership function related to time, and this second membership function is used to perform fuzzy inference on the control rule and the internal state transition rule. A fuzzy control method characterized by evaluation by calculation.