JPH09105652A - Automatic sensory inspection method for articles - Google Patents

Abstract

(57) Abstract: When an article is subjected to a sensory inspection to determine whether it is good or bad, when a sensory characteristic curve based on analog data detected by a sensory sensor is simply binarized to determine whether it is good or bad, it is easily affected by external disturbances. I made an erroneous decision due to noise caused by vibration,
This is the point that the abnormality of the article cannot be discriminated from the noise and the reliability of the inspection is lowered. When determining the quality of an article by a sensory test, an image of a sensory characteristic curve A of the article based on analog data detected by a sensory sensor is image-processed, and the coordinates are divided in the XY axis directions by the number of divisions according to the image resolution. And digitizing the analog data with each other, and comparing the XY coordinate data of the digitized analog data with the corresponding reference data of non-defective product, the degree of coincidence between the two data is determined by fuzzy inference, and all the coordinates are determined. The quality of the above-mentioned sensory characteristic curve A is determined by the determination over time, and the quality of the article is also determined.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an automatic sensory inspection method for articles, which is used to judge the quality of the article by sensory inspection.

[0002]

2. Description of the Related Art There is known a technical means for sensory-testing an article such as a product to determine whether the article is a good article or a defective article. The above-mentioned judging means judges whether the article is good or bad by performing a sensory test by a sensory sensor such as a vibration, sound pressure or load sensor to detect analog data, and then binarizing it by a controller with a built-in comparator. is there. For example, as shown in FIG. 4, a sensory characteristic curve (G) (NG) of an article based on analog data detected by a sensory sensor is binarized into a high and low digital signal at a threshold value (T). Therefore, as shown in the characteristic curve (G), if no portion exceeding the threshold value (T) is detected, it is determined as good product data, while as shown in the characteristic curve (NG), the threshold value (T) is determined. If a portion (Na) (Nb) that exceeds is detected, it is determined as defective product data.

[0003]

The problem to be solved is that, when the sensory characteristic curve based on the analog data detected by the sensory sensor is simply binarized to determine whether it is good or bad, it is easily affected by disturbance, and noise due to external vibration causes This is a point in which an erroneous determination may be made, for example, a non-defective product may be determined to be a defective product, and the product abnormality and noise cannot be discriminated, and the reliability of the inspection decreases. Therefore, it is an object of the present invention to provide an automatic sensory inspection method for an article that is less susceptible to the influence of disturbance when it is judged whether the article is good or bad by a sensory test, by discriminating between an abnormality of the article and noise.

[0004]

According to the present invention, in determining the quality of an article by a sensory test, the sensory characteristic curve of the article based on the analog data detected by the sensory sensor is image-processed, and the number of divisions is determined according to the image resolution. A step of digitizing the analog data by dividing coordinates in the XY axis directions, and XY of the digitized analog data
Each coordinate data is compared with the corresponding non-defective reference data to determine the degree of coincidence of both data by fuzzy inference, and the quality of the sensory characteristic curve is determined by the determination over all the coordinates,
In addition, the method further includes the step of determining the quality of the article.

[0005]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of an automatic sensory inspection method for articles according to the present invention will be described below with reference to FIGS. 1 (a) and (b) to FIG. First, as shown in FIG.
A sensory sensor such as sound pressure or load automatically performs sensory inspection of the article to detect analog data, thereby drawing a sensory characteristic curve (A) of the article. Next, as shown in FIG. 1B, the characteristic curve (A) is image-processed by the image processing device,
The sensory characteristic curve (A) based on analog data is digitally divided by the number of divisions according to the image resolution, for example 512, in the XY axis directions (where the X direction is the horizontal axis and the Y direction is the vertical axis). Turn into. Therefore, for each of the divided coordinates, the XY coordinate data (Xn, Yn) of the characteristic curve (A) (where n = 1
Take out ~ 512). Then, in the X coordinate data (Xn), the Y coordinate data (Yn) indicating the sensory characteristic value of the article is compared with the preset reference data (An) of the good product in the same X coordinate data (Xn). Difference of (Qn = Yn-
An) is calculated. Alternatively, the difference (Qn) is used as the reference data (An)
It is also possible to divide by and calculate the percentage value of the difference (Qn) for standardization. Therefore, X is calculated for each coordinate based on the difference (Qn).
The degree of coincidence (Vn) between the Y coordinate data (Xn, Yn) and the reference data (Xn, An) is determined by fuzzy inference, for example, by calculating the center of gravity. Further, the quality of the sensory characteristic curve (A) is determined by the fuzzy determination over all 512 XY coordinates, and the quality of the good or defective article is also determined.

At this time, if the difference (Qn) between the XY coordinate data (Xn, Yn) of the sensory characteristic curve (A) and the reference data (Xn, An) is small, the algorithm in the above fuzzy reasoning has a degree of coincidence ( Vn) becomes large and the sensory test characteristic curve (A) of the article approaches the preset reference data curve of non-defective product. Also, if the difference (Qn) is large, the degree of coincidence (Vn) becomes small and the reference data curve It should deviate from it. Therefore, the following fuzzy rules are created according to the above algorithm.

(I) IF Qn (n = 1 ...) = ZR (Zero), THEN
Vn (n = 1 ...) = PL (Positive Large) (II) IF Qn (n = 1 ...) = PS (Positive Small), THEN V
n (n = 1 ...) = PS (III) IF Qn (n = 1 ...) = PM (Positive Medium), THEN
Vn (n = 1 ...) = ZR As a membership function for executing the fuzzy rule, triangular membership functions (Ma) (Mb) are set as shown in FIG. The membership function (Ma) is for the input part (% value), and the membership function (Mb) is for the output part (coincidence). Therefore, for example, if Qn = 5% as the input data, the goodness of fit is 0.5 for rule (I), 0.5 for rule (II), and 0 otherwise. Therefore, output by the center of gravity calculation (coincidence)
(Vn) is around 2 and is quite close to the reference data.
The above calculation is performed over all XY coordinates to determine the quality of the sensory characteristic curve (A), and also determine the quality of the article.

Further, in fuzzy reasoning, there is a determination means based on probability in addition to the determination based on the above-described center of gravity calculation. For example, FIG.
As shown in, X coordinate data (Xn) (however, n = 1 to 512)
Membership function (Mc) of fuzzy set of Y coordinate data (Yn) in (where (ZRa) is reference data, (PS
a) (NSa) is each position shift fuzzy set}, and 512 determination probabilities (Dn) are set for each coordinate. Therefore, for example, the reference data of the Y coordinate data (characteristic value) (Yn) in the X coordinate data (Xn) and each suitability (Ra) (Rb) for the positional deviation are detected. And the goodness of fit (Ra)
Is larger and the smaller the fitness (Rb) is, the closer to the reference data. Therefore, they are compared with the determination probability (Dn). For example, when Ra>Dn> Rb, Y coordinate data (Yn), that is, The characteristic value is determined to be normal, and the determination work is performed 512 times for each coordinate. Therefore, for example, the number of normal determinations or the product of the goodness of fit (Ra) (Rb) of all characteristic values or the like is determined as a determination criterion from all the 512 determination results, and the quality of the sensory characteristic curve (A) is determined in the same manner as above. The quality of the article is also determined.

[0009]

According to the present invention, when the quality of an article is judged by the sensory inspection, the sensory characteristic curve based on the analog data detected by the sensory sensor is image-processed, and the number of lines corresponding to the image resolution, for example, XY is used. The coordinates are divided in the axial direction and digitized, and each XY coordinate data of the digitized analog data is compared with the corresponding reference data of a non-defective product, and the degree of coincidence between the two data is determined by fuzzy inference, and determination is made over all the coordinates. Judge the quality of the sensory characteristic curve more,
In addition, since the quality of the article is determined, it is difficult to be affected by disturbance, and it is possible to distinguish between the article abnormality and the noise and there is no erroneous determination,
Further, since the sensory characteristic curve is divided into 512 in the XY axis directions and the data is taken out, the reliability of the inspection is improved.

[Brief description of the drawings]

FIG. 1A is a graph of a sensory characteristic curve of an article by analog data showing an embodiment of an automatic sensory inspection method for an article according to the present invention. FIG. 1B is a graph showing digital division of the graph of FIG. 1A in the XY axis directions.

FIG. 2 is a waveform diagram of each membership function of the input / output unit for performing fuzzy inference of the automatic sensory inspection method for articles according to the present invention.

FIG. 3 is a waveform diagram of another membership function for implementing fuzzy inference in the automatic sensory inspection method for articles according to the present invention.

FIG. 4 is a graph of a sensory characteristic curve of an article by analog data showing an example of a conventional automatic sensory inspection method for an article.

[Explanation of symbols]

 A sensory characteristic curve

Continuation of front page (51) Int.Cl. ⁶ Identification code Office reference number FI Technical display area G06F 15/70 455A

Claims (1)

[Claims] 1. When determining the quality of an article by a sensory test, the sensory characteristic curve of the article based on the analog data detected by the sensory sensor is image-processed, and the coordinates are divided in the XY axis directions by the number of divisions according to the image resolution. And digitizing the analog data with each other, and comparing the XY coordinate data of the digitized analog data with the corresponding reference data of non-defective product, the degree of coincidence between the two data is determined by fuzzy inference, and all the coordinates are determined. A method of automatically sensory inspecting an article, comprising the step of determining pass / fail of the above-mentioned sensory characteristic curve through determination over time, and also determining pass / fail of the article.