JPH0371303A - Fuzzy controller - Google Patents

Abstract

PURPOSE:To obtain the fuzzy controller for executing a fuzzy inference by a rule of the necessary minimum by eliminating a rule having no variation quantity of an output variable value by the inference, and a rule scarcely having the variation quantity, as a rule for exerting no influence on the inference, from among fuzzy rules which are stored, based on an operation command of a rule evaluation. CONSTITUTION:The fuzzy controller is provided with an inference control means 2 for executing an inference by other rule than a specific rule from among fuzzy rules stored in a storage means 3, based on an operation command of a rule evaluation, a variation quantity calculating means 2 for calculating the variation quantity to a usual output variable value of an output variable value by its inference, and an informing means 4 of the variation quantity. In this state, the variation quantity of the output variable value is calculated, and the rule having no variation quantity, and the rule scarcely having the variation quantity are eliminated as the rule for exerting no influence on the inference. In such a way, the fuzzy control whose processing operation is executed quickly can be obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a rule evaluation configuration for a fuzzy controller.

(Prior art) In a fuzzy controller, the degree of fitness for each antecedent membership function of a given input variable value is calculated, and the antecedent degree of fitness, which is the minimum value of these degrees of fitness, is calculated for each rule. The consequent membership function is truncated in each rule based on the antecedent fitness, and then the consequent membership functions that have been truncated are superimposed, and the determined value corresponding to the center of gravity is calculated. It is outputting for the input variable value.

(Problem to be Solved by the Invention) As described above, in the fuzzy controller, inference is performed for input variable values using all fuzzy rules. Therefore, when inference is made by software, the inference is performed sequentially, so if there are many rules, the processing time becomes longer, and output variable values cannot be obtained quickly.
When inference is performed by hardware, an inference board corresponding to the number of rules is required, which poses the basic problem of making the device complex and expensive.

On the other hand, fuzzy rules are set by those involved in controlling the equipment to be controlled, taking into account control know-how, etc., and some fuzzy rules actually have little effect on the output. In many cases, it is included.

The present invention was made in view of the above problems, and an object of the present invention is to provide a fuzzy controller that can easily recognize fuzzy rules that do not affect output.

(Means for Solving the Problems) In order to achieve such an object, the present invention calculates output variable values for given input variable values based on fuzzy rules and membership functions stored in a storage means. In a fuzzy controller that infers and outputs an inference by an inference means, based on an operation command for rule evaluation, the inference causes the inference means to make an inference in a rule excluding a specific rule from fuzzy rules stored in the storage means. a control means, and a change amount calculation means for calculating an amount of change in an output variable value based on the inference of the inference means in a rule other than the specific rule, with respect to a normal output variable value;
and a means for notifying the amount of change.

(Operation) According to the above configuration, based on the operation command for rule evaluation,
Inference is performed using the fuzzy rules stored in the storage means excluding specific rules, and the amount of change in the output variable value based on the inference from the normal output variable value is calculated, and the amount of change is reported. be done.

Then, rules with no amount of change or rules with a small amount of change are removed as rules that do not affect inference.

(Example) Hereinafter, an example of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing the overall configuration of the fuzzy controller of the present invention, and shows a type in which the inference process is processed by software.

In the figure, reference numeral 1 denotes a control section, and the control section 1 is composed of a CPU 2 constituted by a microcomputer and a memory 3. A CRT 4 and a keyboard 5 as an operation section are connected to the control section 1, respectively.

In the above configuration, the CPU 2 of the control unit 1 performs inference operations as a fuzzy inference means, and also controls the operation of the CRT 4. Based on the operation command for rule evaluation from the keyboard 5, the CPU 2 stores information in the memory 3. An inference control means that performs inference using a rule excluding a specific rule from fuzzy rules, and a change amount calculation that calculates the amount of change in the output variable value from the normal output variable value due to inference in the rule excluding the specific rule. It also functions as a means.

The memory 3 as a storage means stores a basic operation program of the device, a fuzzy inference program including various membership functions and various fuzzy inference rules, and the amount of change obtained by calculation.

The CRT 4 also functions as a means for notifying the amount of change.

The input of the membership function and fuzzy inference rules is
This is done while checking the display on the CRT 4 using the keyboard 5, and the membership functions and fuzzy rules based on the input values are stored in the memory 3.

Next, the operation of the fuzzy controller will be explained.

The fuzzy controller controls the sensor i of the device 10 to be controlled.
If there is human power from each of them, the CPU 2 performs fuzzy inference based on the membership function and fuzzy inference rules stored in the memory 3, and outputs the output variable value to the control target section 12 of the control target device 10 as a manipulated variable. It is something to give.

Next, the rule evaluation operation in the fuzzy controller of the present invention will be explained with reference to the flowchart in FIG.

First, when an evaluation operation is performed on the keyboard 5 (step S1), [ir X1=A
, X2=B then y=C], and the output variable values thereof are calculated (step S2). This output variable value is compared with the normal output variable value in the input variable value-output variable value correspondence table T stored in the memory 3, the amount of change is calculated (step S3), and the amount of change is stored in the memory 3. (Step S4).

The above input variable value-output variable value correspondence table T is created before this fuzzy controller is used for control.
Output variable values U for all combinations of human variable values XI and X2 are fuzzy inferred, and the output variable values are stored in the memory 3 in matrix form as shown in FIG. For example, input variable values xi are i, x
When 2 is j, the output variable value is Uij.

Then, it is determined whether the evaluation of all rules has been completed (step S5), and if the evaluation has not been completed, the process moves to step 2 to evaluate the next rule. Evaluation takes place. When the evaluation of all the rules is completed, the total amount of change stored in step 4 is displayed on the CRT 4 in the output variable value change amount diagram G in a matrix format as shown in FIG. 4 (step S6). ).

This display is performed in accordance with the degree of the amount of change, and is displayed in shading as shown in FIG. 4, for example, depending on the magnitude of the amount of change, and the amount of change can be recognized by the degree of the display. For example, in the display example of Fig. 4, xl with no display is PL
There is no change in all rules where X2 is PL and NL, and the amount of change increases as the display density increases, x 1 = ZR, X 2 = PS
This shows that the amount of change in the rule is the largest.

Therefore, if there is a display like the one above, it can be determined that all the rules that set XI as PI and all the rules that set x2 as PL and NL have no effect on the output. Just delete the rule. The amount of change may be displayed by color, or may be displayed numerically.

The calculation of the amount of change in the above is, for example, xl=NS, x
The range in which the output variable value changes when the rule that sets 2=PS is removed is only the part corresponding to the membership functions NS and PS related to that rule, so that part, that is, xl is a to b, This is done by calculating the sum of output variable values in the part where x2 is c-d, and comparing that sum with the normal sum of output variable values.In this way, the overall output variable value is calculated for each rule. If the amount of change is calculated in the necessary portion instead of the above, the calculation process can be performed quickly.

In the above embodiment, the amount of change is notified using the CRT 4, but the amount of change may be printed out using a printer.

(Effects of the Invention) Therefore, according to the present invention, based on the operation command for rule evaluation, inference is performed on rules excluding a specific rule from among the fuzzy rules stored in the storage means, and an output based on the inference is performed. The amount of change in the variable value from the normal output variable value is calculated, and the amount of change is reported.
According to the notification results, rules with no amount of change or rules with a small amount of change can now be removed as rules that do not affect inference.

As a result, it has become possible to obtain a fuzzy controller that performs fuzzy inference using the minimum necessary rules. A fuzzy controller with simple configuration and low cost can now be obtained.

[Brief explanation of drawings]

Figures 1 to 4 relate to embodiments of the present invention; Figure 1 is a configuration diagram of a fuzzy controller, Figure 2 is a flowchart for explaining the operation, and Figure 3 is a correspondence table between input variable values and output variable values. 1 and 4 are diagrams showing examples of display of the amount of change. 2...CPU (mountain means) 3...Memory (4...CRT (inference means, storage means) notification means) Inference control means, change amount field

Claims (1)

[Claims] (1) In a fuzzy controller that infers and outputs an output variable value for a given input variable value by an inference means based on fuzzy rules and membership functions stored in a storage means, based on an operation command for rule evaluation. , an inference control means for causing the inference means to make an inference based on a rule excluding a specific rule from among the fuzzy rules stored in the storage means; and an output based on the inference of the inference means in the rule excluding the specific rule. A fuzzy controller comprising: a change amount calculation means for calculating a change amount of a variable value with respect to a normal output variable value; and a change amount reporting means.