JP2672641B2 - Pass / fail judgment device based on subjective evaluation value - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Field of Industrial Application The present invention relates to a pass / fail judgment device based on a subjective evaluation value for making a comprehensive evaluation by fuzzy inference using evaluations of a plurality of items as input values. .

(B) Conventional Technology As an apparatus for subjectively evaluating an article as a non-defective article on behalf of a human, there is an automatic sensory inspection apparatus disclosed in JP-A-61-23966. In this device, the tapping sound of the sample is captured by a microphone, and the level of each frequency of the tapping sound is compared with the levels of the tapping sounds of the non-defective product and the defective product that are stored in advance, and the quality of the sample is determined. As a result, it is possible to accurately and easily perform sample selection, which was conventionally performed based on experience and know-how.

(C) Problems to be Solved by the Invention However, in the inspection of articles, there are some things that should be judged based on the subjective criteria of human beings, and the evaluation items thereof may be various, and it is important that a plurality of evaluation items be important. The degree is not constant, and it is not clear which evaluation item should be emphasized. The above-mentioned conventional device has a problem that it is not possible to evaluate the inspection result based on the above-mentioned subjective criteria, since the inspection result only determines whether or not the inspection pattern matches a standard pattern stored in advance. Moreover, it was not possible to easily modify or change the evaluation criteria.

An object of the present invention is to make a comprehensive evaluation by performing fuzzy inference using evaluations of each of a plurality of items as inputs, and based on a subjective evaluation value that can make an evaluation based on an individual's subjective criteria. It is to provide a pass / fail judgment device. Another object of the present invention is to provide a sensory test device that can easily correct and change the criteria for processing subjective evaluation values.

(D) Means for Solving the Problem The apparatus of the present invention is a presentation means for sequentially presenting a plurality of preset evaluation item contents for determining acceptance / rejection for one evaluation object to the operator as input auxiliary information. , Input means for the operator to manually input the subjective evaluation value for each evaluation item content in response to each presentation of the input auxiliary information by the presenting means, and fuzzy knowledge stored in the replaceable storage medium. Fuzzy knowledge reading means for reading, and a fuzzy knowledge obtained by integrating a plurality of subjective evaluation values input by the input means by fuzzy inference based on the fuzzy knowledge read by the fuzzy knowledge reading means. It is characterized by comprising an inference means and a display means for displaying the pass / fail judgment result of the evaluation object according to the definite value output by the fuzzy inference means.

(E) Operation In the present invention, fuzzy inference is performed by a plurality of evaluation values input from the input means, and a comprehensive evaluation is made.

As is well known, the fuzzy inference means is composed of a fuzzy arithmetic unit for performing fuzzy arithmetic and a defuzzy phi unit for performing fixed value arithmetic. The fuzzy operation unit is equipped with a membership function generator according to a predetermined fuzzy rule, calculates membership values for input variables, and infers the calculated inference value based on the result to the defuzzy unit. Output. This fuzzy rule is expressed in the form of if (x ₁ = A and x ₂ = B ...) Then (y = Z) ... Formula ₁ (x ₁ = A and x ₂ = B ...) Section,
(Y = Z) is called the consequent.

FIG. 8 is a diagram illustrating a known method of outputting an inference result according to the above fuzzy rule. (A),
(B) shows the membership function corresponding to the _two variables (x ₁ , x ₂ ) in the antecedent part that is the input value, and FIG. 6C shows the membership function that corresponds to the consequent part that is the output value. Represents
Here, two membership functions of the antecedent part are shown, but if the types of variables (number of terms) of the antecedent part increase, the membership function also increases accordingly. In each figure, the horizontal axis represents the value of the variable, and the vertical axis represents the position of membership (degree of membership).

Now, the value of the variable x ₁ of the first item of antecedent is assumed to be x ₁ ', the appertaining at that time is 0.5 (see FIG. (A)). If the value of the variable x ₂ of the second item of the antecedent part is assumed to be x ₂ ', the appertaining at that time is 0.3 (see FIG. (B)). In such a case, the fuzzy calculation unit takes the smallest value in each degree of belonging. That is,
In the above example, the degree of affiliation of 0.3 is selected. Next, the membership function corresponding to Z is truncated at the above-mentioned affiliation degree 0.3, and the center of gravity y 'of the lower trapezoid S is obtained. Then, it is output as the inference result of this y '.

The above inference is performed for one rule, but generally, a plurality of rules are set. In this case, the inference result shown in FIG. 8C is output for each rule.
Then, the trapezoidal portion output for each rule is logically ORed, and the center of gravity y ″ of the logically ORed portion (hatched area in FIG. 8D) is output as the final value of the logic. 8
The output value is obtained so that the input value of the membership function of FIGS. (A) and (B) takes an intermediate value.

In the above logical method, the logical product operation of the degree of belonging to the antecedent part (the operation that chooses the smaller one) and the logical sum operation rule of the trapezoidal part and the min-max rule for the consequent part are called Executed in the circuit and the consequent OR circuit. In the present invention, the center of gravity y ″ in FIG. 8 (D) is output as the comprehensive evaluation.

Further, if a means for reading the stored contents of the card medium is provided, the aforementioned fuzzy inference is performed according to the fuzzy rules and membership functions written in the card medium. Further, the portable type is realized by supplying the power source with the battery or storing the power source in the case.

(F) Example FIG. 1 is an external view of a sensory test apparatus which is an example of a pass / fail judgment apparatus based on a subjective evaluation value according to the present invention.

On the front of the sensory test apparatus main body 1 in which a control unit and the like are housed in a portable case, a display unit 2 for displaying input items and the like and an input key 3 are provided in the center. Display 2
The contents of a plurality of evaluation items for which evaluation values are to be input are displayed in order by operating the selection key 4. After inputting the evaluation value for each item using the input key 3 and the setting key 5,
When the processing start key 6 is operated, the quality of the object to be evaluated is comprehensively evaluated by fuzzy inference. The result is lamps 7-9
And is displayed by the buzzer 10. A card 12 is detachably attached to one side surface of the sensory inspection apparatus body 1. The card 12 stores evaluation item data, fuzzy rule data, membership function data, etc., which will be described later.

 FIG. 2 is a block diagram of the sensory inspection apparatus.

A ROM 23 and a RAM 24 are connected to the CPU 21 via an internal bus 22. The following programs for controlling the input / output devices and programs for processing related to fuzzy inference are written in the ROM 23 in advance. A predetermined memory area of the RAM 24 is assigned to the working area of the CPU 21. The CPU 21 has a card interface section 25, a key unit interface section 26, a lamp interface section 27,
A buzzer interface section 28 and a display interface section 29 are connected.

The card interface unit 25 includes a card reader for reading the data written in the card 12 inserted in the sensory test apparatus body 1. The lamp interface section 27 includes drivers for the pass lamp 7, the unknown lamp 8 and the fail lamp 9. The buzzer interface unit 26 includes a driver that drives the buzzer 10. The display interface 29 includes a driver that controls lighting of the display 2.

The battery 30 is detachably attached to the sensory inspection apparatus body 1 described above, and when the power switch 11 provided on the upper front surface of the sensory inspection apparatus body 1 is turned on, power is supplied from the battery 30 to each part in the apparatus. To be done. When the battery 30 can be loaded in the sensory inspection apparatus main body 1 in this manner, the sensory inspection apparatus can be configured to be portable.

FIGS. 3 to 6 are views showing the recorded contents of the card which can be inserted into and removed from the sensory inspection device, and FIG. 3 shows evaluation item data showing the contents of items to be evaluated, and FIG. Shows fuzzy rule data, and FIGS. 5 and 6 show membership function data, which are data used for fruit evaluation.

When the evaluation setting is input, the contents of the evaluation item data shown in FIG. 3 are displayed one by one on the display unit 2. The display contents are switched by operating the selection key 4, and one of the M items is displayed alternatively.

The fuzzy rule data shown in FIG. 4 represents if-then type fuzzy rules. That is, fuzzy rule data the uppermost represent _{if (V 11 = L 11 and} V 12 = L 12 and V 13 = L 13 ... and V 1M = L 1M) then W = Y 1. That is, V _{11 to} V _1M correspond to the respective terms in the antecedent part of the first equation, and the number of terms is equal to the number of items shown in FIG. Thus, one line in FIG. 4 represents one fuzzy rule, and represents N fuzzy rules as a whole.

As shown in FIG. 5, in the antecedent part, there are membership functions as many as the number of terms M multiplied by the type of label. Now, assuming that the number of labels is seven, any of the labels "A" to "G" shown in FIG. 5 is applied to each item of the antecedent part. That is, in FIG. 4, each of the labels (L ₁₁ , L ₂₁ ... L _N1 ) of the first term of the antecedent part of each fuzzy rule is labeled with the label “A of the membership function shown in the leftmost column of FIG. Any of "~" G "applies. Similarly, each of the labels (L _{21 to} L _N2 ) of the second item of the antecedent part applies to one of the labels of the membership function in the second column from the left side of FIG. Is one of the labels represented by the membership function shown in the rightmost column of FIG. Further, any of the labels "A" to "G" of the membership function shown in FIG. 6 is assigned as the consequent parts Y _{1 to} Y _1n of the fuzzy rule shown in FIG. The fourth above
The fuzzy rules and membership functions shown in FIGS. 6 to 6 are empirically determined in advance.

FIG. 7 is a flowchart showing a processing procedure of a CPU which constitutes a part of the sensory inspection apparatus.

When the power switch 11 is turned on, the CPU 21 initializes the display on the display unit 2 (n1) and waits for the insertion of the card 12 (n1).
2). When the card 12 is inserted, the data written in the card 12 is read and stored in the RAM 24 (n
3). Then, the processing start screen is displayed on the display 12 (n
4), the evaluation item counter i is set to 0 (n5). The evaluation item counter i is assigned to a predetermined memory area of the RAM 24 and is a counter for specifying any of the evaluation items shown in FIG.

Further, all the flags indicating the input completed state of each evaluation item are reset, and the evaluation content for the evaluation item 1 selected by operating the selection key 4 is displayed on the display device 2 (n7-
n12). The evaluation value for each item input according to the display contents of the display 2 is stored in a predetermined memory area of the RAM 24, and the flag corresponding to that item is set (n14). When the evaluation value is input, membership values (affiliation values) for all propositions that process the evaluation value as a fuzzy variable are calculated (n15).

When the evaluation values for all evaluation items are input by the processing of n16 and n17, the suitability of the antecedent part for each of the fuzzy rules shown in FIG. 4 is calculated by executing the mini-max rule (n20). . Next, the membership function for each fuzzy rule is calculated by executing the mini-max rule (n21, n22). N for all fuzzy rules
When the processing of 16 to n22 is completed (n23, n24), the defuzzification of the membership function of each fuzzy rule is executed,
A definite value e is generated (n25). The lamps 7 to 9 and the buzzer 10 are driven according to the value of the definite value e (n26).

Through the above processing, fuzzy inference is performed using the evaluation values for each of the evaluation items shown in FIG. 3 as input values.
The comprehensive evaluation can be obtained by a fuzzy rule and a membership function that are determined based on experience in advance.

In this embodiment, the membership function has a relationship that can be expressed by mathematical expressions as shown in FIGS. 5 and 6, but in each membership function, the numerical value corresponding to each evaluation value of each evaluation item is set. May be set randomly based on experience.

(G) Effect of the Invention According to the present invention, there is an advantage that a comprehensive evaluation can be made according to a standard based on an individual subjective evaluation. Moreover, if the operator inputs a subjective evaluation value for each evaluation item while simply referring to the input assistance information without paying attention to the complicated criteria of the overall evaluation, a predetermined fuzzy logic is automatically set. A comprehensive evaluation that combines those subjective evaluation values based on knowledge is performed, and a pass / fail judgment result is obtained based on that, so that comprehensive evaluation work based on multiple evaluation items can be performed easily and stably. There is an effect that the evaluation work becomes efficient and the personal effect of the pass / fail judgment is reduced.

In addition, by reading the fuzzy rules and membership functions of the fuzzy reasoning related to the comprehensive evaluation as data from the removable card, it is possible to easily correct or change the evaluation criteria, and to perform various subjective evaluation tasks. It has the advantage of being able to respond appropriately. Also, by supplying power from a battery or storing it in a case, it becomes portable, which makes it very convenient.

[Brief description of the drawings]

FIG. 1 is an external view of a sensory inspection apparatus according to an embodiment of the present invention, FIG. 2 is a block diagram of the sensory inspection apparatus, FIG. 3 is a diagram showing evaluation items in the sensory inspection apparatus, and FIG. The figure which shows the fuzzy rule in the fuzzy inference of the inspection device, 5 and 6 is the figure which shows the membership function of the fuzzy inference in the same sensory inspection device, Figure 7 is the CPU which forms a part of the same sensory inspection device 5 is a flowchart showing the processing procedure of step S1. 8 (A) to 8 (D) are diagrams for explaining a known fuzzy inference method. 1 ... Sensory test device main body, 2 ... Display (display means), 3-6 ... Key (input means), 12 ... Card, 25 ... Card interface section.

Claims (1)

(57) [Claims] 1. Presentation means for sequentially presenting a plurality of preset evaluation item contents for determining acceptance / rejection for one evaluation object to the operator as input assistance information, and each presentation of the input assistance information by the presentation means. According to the above, the operator manually inputs a subjective evaluation value for each evaluation item content, a fuzzy knowledge reading means for reading fuzzy knowledge stored in a replaceable storage medium, and the fuzzy Based on the fuzzy knowledge read by the knowledge reading means, a plurality of subjective evaluation values input by the input means are integrated by fuzzy inference to obtain a definite value, and a fuzzy inference means outputs the fuzzy inference means. A pass / fail judgment device based on a subjective evaluation value, comprising: a display unit that displays the pass / fail judgment result of the evaluation target according to a fixed value.