JP2571145B2 - Control method of plant etc. by fuzzy inference - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to information processing such as control in plants and other large-sized devices.

2. Description of the Related Art Today, various types of information processing based on fuzzy inference called an expert system are performed.

For example, in order to perform process control by this fuzzy inference, the deviation of the current control amount, which is the state of the device to be controlled, from the target value, the amount of change in the deviation, and the like are defined as a fuzzy set as a membership function of the antecedent. , And
Determine the consequent membership function that defines the fuzzy set of manipulated variables, determine the grade of the measured value by associating the observed value (measured value) with the antecedent member function,
According to the value of this grade, the membership function of the consequent part is α
-Perform fuzzy inference so as to cut, and calculate the operation amount by performing defuzzification operation by calculating the center of gravity of the fuzzy set of the consequent part after α-cut, and calculate the operation amount, and ideally the operating state of the device to be controlled. It is controlled to approach the state.

Further, in order to more accurately perform the control and the judgment by the fuzzy inference, the responsiveness of the control by the inference device is improved (for example, Japanese Patent Application Laid-Open No. 63-113.733), and the noise contained in the measured value is reduced. Removal (for example, JP-A-1-119.803)
Have been tried.

[Problems to be Solved by the Invention] Performing fuzzy inference for a part of arithmetic for control and information processing in a large-scale device has been already performed in many fields, but control of a plant and the like and complicated information are performed. In processing, the number of control rules increases, and it becomes necessary to superimpose a large number of rules.

However, when a large number of rules are superposed based on fuzzy inference and inference is performed, an unexpected inference result is calculated, and there is a disadvantage that it is difficult to obtain an accurate control and a determination result.

Means for Solving the Problems The present invention calculates a conclusion fuzzy set by each fuzzy rule based on fuzzy inference, combines the conclusion fuzzy sets by each fuzzy rule, derives an output fuzzy set, and outputs an output fuzzy set. , A first output value is obtained by performing a defuzzification operation based on the area method or the centroid method, and a second output value is obtained by defuzzifying the output fuzzy set also based on the fuzzy entropy method. Is compared with the second output value, and when the difference output is less than a predetermined threshold, based on the first output value or the second output value,
When the difference output is equal to or larger than the threshold, the control is always performed based on the second output value.

In some cases, an output value is simply obtained from an output fuzzy set based only on the fuzzy entropy method.

[Operation] Since the present invention obtains an output value based on the fuzzy entropy method when performing the defuzzification operation on the output fuzzy set, a fuzzy set having a small grade is obtained as a fuzzy set in the conclusion part constituting the output fuzzy set. In the case where there is, it is possible to obtain an output value in which the weight of the conclusion fuzzy set having the small grade is increased.

When performing two kinds of operations, such as an area method and a fuzzy entropy method, when performing a defuzzification operation from an output fuzzy set, a first output value and a second output value which are two types of operation results Is large, the second output value based on the fuzzy entropy method is adopted.
When a fuzzy set having a small grade is included in the fuzzy set of the conclusion part constituting the output fuzzy set, an output value in which the weight of the fuzzy set of the consequent part having the small grade is increased can be obtained.

Embodiment An embodiment of the present invention is a method of deriving one control output based on a plurality of fuzzy rules. As shown in the flowchart of FIG. 1, inference by the area method or the centroid method is performed in the defuzzification operation. To calculate the output value δ, and also calculate the output value ω0 by performing inference by the fuzzy entropy method using fuzzy entropy as a parameter.

That is, parameters such as the antecedent part membership function and the consequent part membership function and the conversion coefficient of the measurement value for the measurement target are set according to the equipment such as the plant, and the measurement values from various sensors are set in the same manner as before. When input as a fact, the control signal is output to the actuator by performing fuzzy inference such as mini-operation or α-cutting the membership function of the consequent part, and the fuzzy inference takes into account the fuzzy entropy inference. Is what you do.

This fuzzy entropy method measures the degree of ambiguity of the conclusion part fuzzy set by fuzzy entropy, synthesizes the conclusion part fuzzy sets obtained individually by inference based on each fuzzy rule, calculates the output fuzzy set, and then outputs The fuzzy set is interpolated as a parameter when performing defuzzification operation. As shown in FIG. 2, this fuzzy entropy is 0 when the grade of the fuzzy set is 0 or 1, and becomes 1 when the grade is 0.5, that is, when the degree of ambiguity is the highest, the Shannon function S
(X) is used.

In this embodiment, which also takes into account the fuzzy entropy method, in order to derive one output from a plurality of fuzzy rules, first, inference is performed with each fuzzy rule. Similarly, the grade of the membership function of the antecedent part is mini-synthesized, and α-cut of the membership function of the antecedent part is performed to obtain the fuzzy set of the conclusion part.

An example of this fuzzy inference is IF a is X and b is Y THEN ω is
When a fuzzy set with membership functions μX, μY, and μB is defined, the grade of the first input a is obtained as shown in FIG. The grade of the input b is obtained, and the fuzzy set A of the conclusion part obtained by cutting the membership function of the consequent part in the grade of the small value in the grade of the first input and the grade of the second input is obtained, or as shown in FIG. As shown, a consequent part fuzzy set A is obtained by compressing the consequent part fuzzy set B to a small grade value in the antecedent part input. If the number of fuzzy rules is n, first the n consequent fuzzy sets are obtained. The partial fuzzy sets A1, A2, A3,..., An are obtained.

Then, the fuzzy sets of the conclusion parts are combined as shown in FIG. 5 to obtain an output fuzzy set, and then a defuzzification operation is performed to calculate an output value. The value δ1 of the area center β of the set is set as the first output value δ, or the center of gravity γ of each conclusion part fuzzy set constituting the output fuzzy set as shown in FIG.
The first output value δ is obtained by calculating the output value δ0 by combining 1, γ2.

As described above, the first output value δ obtained by defuzzifying the output fuzzy set by the area method or the centroid method is calculated, and the second embodiment outputs the second output value ω by the fuzzy entropy method.
0 is also calculated.

This fuzzy entropy method uses the definition of Younger when each membership function of each of the n conclusion part fuzzy sets Aα (α = 1, 2,... N) is μAα (α = 1, 2,... N). And the entropy of each fuzzy set d (A
α), And each representative point corresponding to each barycenter of each conclusion part fuzzy set Aα is represented by ωα. It is calculated by:

Note that μα represents a membership function of the complement set α of the fuzzy set Aα.

The second output value ω0 having the fuzzy entropy as a parameter is a value in consideration of the intermediate grade level, and it is unclear whether or not the measurement fact conforms to the fuzzy set of the antecedent part. Is generated, the output value is obtained by increasing the weight of the conclusion part fuzzy set of the lower grade. Therefore, when the grade of each concluding part fuzzy set is high, the value is close to the first output value δ by the area method or the center of gravity method, but when the concluding part fuzzy set includes a low grade fuzzy set. In this case, the fluctuation range of the output value is suppressed to be small as compared with the area method or the centroid method.

As described above, the first output value δ and the second output value ω0 are obtained, the difference Δω between the first output value δ and the second output value ω0 is further obtained, and the value of the difference output Δω is fixed. Compare to threshold K,
When the difference value Δω is larger than the threshold value K, the judgment is made based on the output value ω0 by the fuzzy entropy method when the difference output Δω is larger than the threshold value K. It is.

Therefore, in this control method, when the difference output Δω, which is the difference between the first output value δ and the second output value ω0, is large, in other words, the degree of conformity to the fuzzy set during inference based on the multiple fuzzy rules is ambiguous. In the case where such information is included, a weight can be added to the ambiguous information amount to narrow the pattern matching range. Then, when the range of the matching is narrowed to some extent, it is possible to switch to the inference method centering on the grade. Therefore, the calculation time for pattern matching can be reduced.

The definition of fuzzy entropy is not limited to the definition of Younger, It is also possible to obtain the entropy of each fuzzy set of the conclusion part based on the definition.

Also, in performing the defuzzification operation using the fuzzy entropy as a parameter, taking into account the width of the base of the conclusion part fuzzy set together with the fuzzy entropy, each conclusion part fuzzy set constituting the output fuzzy set as shown in FIG. Let lα be the width of the base May be used to calculate the second output value ω0.

The threshold value K is set in advance according to the environment to be controlled.

Further, when the difference output Δω, which is the difference between the first output value δ and the second output value ω0, is equal to or larger than the threshold value K, the second output value ω0 is selected. When the difference output Δδ is less than the threshold value K, the second output value ω0 which is similar to the first output value δ is selected without being limited to the case where the first output value δ is selected. Sometimes.

As described above, when the second output value ω0 is always selected and the determination control is performed, the calculation of the first output value δ by the area method or the centroid method is further omitted, and the fuzzy entropy is further reduced. The determination is always made based on the output value ω0 as a parameter, and the tuning and control can be performed by adding weights to the ambiguous information amount while reducing the number of operation processing steps.

[Effects of the Invention] The control method of a plant or the like based on fuzzy inference according to the present invention, when defuzzifying an output fuzzy set obtained by combining the conclusion fuzzy sets by each fuzzy rule,
Performs control and judgment based on the output value ω0 using fuzzy entropy as a parameter, so it is suitable for tuning in an environment in which a trial and error rule or membership function must be given, and inference in an environment where a hypothesis needs to be set. The automatic control setting can be performed by facilitating the integrated development of the system.

Further, in defuzzifying an output fuzzy set obtained by combining the conclusion part fuzzy sets by each fuzzy rule, a first output value based on an area method or a center of gravity method and a second output value using fuzzy entropy as a parameter are calculated. , First
When the difference output Δω, which is the difference between the output value and the second output value, is larger than the threshold value K, the method of performing control and judgment based on the second output provides a trial and error rule or membership function. Suitable for tuning in an inevitable environment, deployment of the inference method in an environment that requires setting of a hypothesis Facilitates integrated thinking, and if the fuzzy set of the conclusion part does not include an ambiguous amount of information, In addition, control can be performed based on grade-based, center-oriented average inference results.

[Brief description of the drawings]

FIG. 1 is a flow chart showing an embodiment of the present invention, FIG. 2 is a graph showing fuzzy entropy, FIGS. 3 and 4 are diagrams each showing fuzzy inference, and FIGS. FIG. 7 is a diagram illustrating defuzzification by an area method and a center of gravity method, and FIG. 7 is a diagram illustrating a fuzzy entropy method.

Claims (2)

(57) [Claims] When obtaining one control output based on a large number of fuzzy rules, each fuzzy set is determined by fuzzy inference based on each fuzzy rule, and each fuzzy set is concatenated with an output fuzzy set. Defuzzification operation is performed from the output fuzzy set by an area method or a centroid method to obtain a first output value δ, and defuzzification operation is performed by an entropy method taking fuzzy entropy into consideration to obtain a second output value. ω0 is determined, and the first output value and the second output value are compared to determine a difference output Δω. If the difference output Δω exceeds a threshold value, the second output value ω0 is used, and the difference output Δω is determined. Is controlled by adopting the first output value δ or the second output value ω0 when the value is less than the threshold value. 2. A method for obtaining one control output based on a large number of fuzzy rules, obtaining each consequent part fuzzy set by fuzzy inference based on each fuzzy rule, combining each consequent part fuzzy set, and outputting an output fuzzy set. Fuzzy inference characterized in that the output fuzzy set is defuzzified to obtain an output value ω0, and the defuzzification operation is performed using an entropy method that takes into account the fuzzy entropy of the fuzzy set of the conclusion part. Control method for plants and so on.