JPH0687046B2 - Steel plate surface flaw inspection method by neural network - Google Patents

Description

TECHNICAL FIELD The present invention relates to a method for inspecting a surface flaw of a steel sheet using a neural network by a reflected light waveform.

[Conventional technology]

A method using reflected light is widely used for inspecting the surface defects of the steel sheet. A typical method is to project a laser beam onto the surface of a steel sheet, photoelectrically convert the reflected light, compare the converted output (voltage waveform) or its differential output with a threshold value, and make the flaw equal to or greater than the threshold value.

As a method of using a method other than the threshold value, Japanese Patent Laid-Open No. 62-278403
There is. In this method, laser light is projected onto the surface of the steel sheet, the reflected light is photoelectrically converted, and the surface roughness is obtained from the average voltage. If the roughness is high, the reflected light becomes scattered light, and the photoelectric conversion output decreases, so if the relationship between the roughness and the photoelectric conversion output is obtained in advance,
From this relationship, the roughness can be known from the photoelectric conversion output.

In addition, JP-A-62-235511 projects a laser beam on an object to be inspected,
The surface state is detected by utilizing the fact that the incident point of the reflected light deviates from the normal state by ΔX in the place where the surface state is abnormal (where there is a surface defect or unevenness).

[Problems to be Solved by the Invention]

In the conventional inspection of steel plate surface flaws, it is general to judge the photoelectric conversion output of the reflected light with a threshold value to determine whether there is a flaw,
In this method, since the flaw is judged by the absolute reflected light intensity, the difference in the degree of reflection between steel types and the influence of formation are not considered. If the threshold value is set low, it is not a flaw but the part where the reflection is a little abnormal is erroneously detected as a flaw (over detection). ) Increases. Therefore, it is necessary to set the threshold value to an appropriate value, but it is difficult to select and adjust the appropriate value because it changes depending on the steel type, formation, etc., and this method has difficulty in determining accuracy. In addition, there is a problem in that if the type of flaw to be detected changes, it cannot respond quickly.

The present invention improves such points, and obtains a feature amount from the photoelectric conversion output of reflected light from the conversion output of a narrow range A including a flaw and a wide range B including the range A. An object of the present invention is to provide a flaw inspection method capable of avoiding over-detection and oversight by making a flaw determination using a circuit network and avoiding the trouble of selecting a threshold value.

[Means for Solving the Problems]

In the present invention, the surface of the steel sheet is scanned with a laser beam, the reflected light is photoelectrically converted, and further A / D converted, and a small range A corresponding to a flaw portion of the steel sheet and a large range including the plurality thereof (steel plate width or A plurality of feature quantities representing the reflected light waveform are calculated from the converted output in the range (one-third range).

The number of nodes in the input layer is the same as that of the plurality of features, and the number of output nodes is one corresponding to the presence / absence of flaws. The neural network is pre-learned. It causes more defects / no output.

There are quadratic average, cubic average, differential range, etc. as the feature amount, but the normalized quadratic average represented by the following formula And the normalized differential range x _DRS is appropriate.

[Operation] In this steel sheet surface flaw inspection, flaws / no outputs can be obtained by calculating the feature amount from the converted output in the small / large ranges A and B, and applying the obtained feature amount to the neural network. There is no problem as to where to set the threshold. In addition, both false detection and over-detection can be made small, and if the threshold value is set high as in the conventional method, false detection will occur frequently, and if the threshold value is set low, over-detection will occur frequently. Never.

〔Example〕

An embodiment of the present invention is shown in FIG. 10 is a steel plate to be inspected, 11 is a laser oscillator, 12 is laser light, and 13 is its receiver. For example, the steel plate 10 moves in the steel plate longitudinal direction indicated by the arrow direction F ₁ and the laser beam reciprocates in the steel plate width direction indicated by the arrow F ₂ , whereby the entire surface of the steel plate 10 is scanned. The receiver 13 comprises an optical fiber and light receiving elements (photomulsions) at both ends thereof, the reflected light of the laser light projected on the steel plate enters the optical fiber, and the photoelectric conversion output thereof is output by the light receiving elements at both ends. This type of receiver also provides scanning position information on the steel plate according to the incident position of reflected light.

The output of the receiver 13 is differentiated by the differentiation circuit 21 and then A / D
A / D conversion is performed by the converter 22. FIG. 2A shows the receiver output, and FIG. 2B is the differential output thereof. B indicates the plate width of the steel plate, and the receiver output rises / falls at the plate edge and makes a recess at the flaw position. The A / D converter 22 samples the differential waveform of (b) at an extremely short cycle (20 to 100 nS) to obtain an analog amount as shown in (c), and converts it into an 8-bit digital value in this example.

The feature amount is calculated from the data thus collected.
The second-order average and the differential range are used as the feature quantity,
The calculation is performed for a small range A that covers a flaw and a wide range B that covers the small range A. In FIG. 2 (a), the range B is the entire plate width, but it may be 1/2, 1/3, etc. of the plate width. 1/2, 1/3 ...
In the case of ..., 2 sets, 3 sets of the apparatus of FIG.
It is advisable to divide the surface of the steel sheet into two, three, and so on, and inspect (parallel simultaneous inspection). The range A is a minute range (for example, 5 mm width) that sufficiently covers a flaw of an expected size, and the range B includes a plurality of ranges A. Second-order average of range A and B ₂ ,
_zall is expressed by the following equation.

Here, y _i is the A / D conversion value at position x _i , x _s and x _e are the range B
I values at the beginning and end of i, i _o and i _o + a are i at the beginning and end of the range A
The values E ₁ and E ₂ are the reciprocals of the number of y in the ranges A and B. range
A, the differential range x _DR of B, x _{DRall is, x DR = ABS (maxy-} miny) ...... (3) i o <y <i o + a x DRall = ABS (maxy-miny) ...... (4) x It is represented by _s <y <x _e . The differential range of the range B is calculated, the standard deviation σ _xDR is calculated from the past data by the following equation, and this is updated each time a new differential range is obtained.

_As a calculation method of σ _xDR , there is a method in which calculation is performed every time in synchronization with calculation of other feature amounts, and calculation is performed using a value obtained from sample data and is a constant for online use. Next, normalize. As shown in the following equations (5) and (6), the equation (1) is divided by the equation (2) to obtain a quadratic average normalized value _2s, and the differential range of the range A is the standard deviation σ of the range B. _Divide by _xDR to obtain the normalized value x _DRS of the differential range of range A. _{2S = 2 / 2all ...... (5} ) x DRS = x DR / σ xDR ...... (6) The normalized feature quantity _2S, performs flaw determination using x _DRS. As shown in FIG. 2 (d), the differential range x _DR and the quadratic average ₂ are plotted on the ordinate and the abscissa on the regions with flaws ○ and flaws ×, as shown in the figure. Is non-linear. As shown in FIG. 2 (e), it is possible to perform a regression analysis if separated by a 45 ° line, but in FIG. 2 (d), a regression analysis is difficult. Therefore, a neural network is used to output defect / absence.

As shown in FIG. 2 (f), this neural network has a three-layer structure of an input layer, an intermediate layer, and an output layer, and the number of nodes is two in the input layer, three in the intermediate layer, and one in the output layer. A bias layer (for learning promotion) is added to. The relationship between the input and output of each node is the sigmoid function a (1 + e ^-x ) ^-1 ,
The output of the bias layer is 1, and the learning rule is the generalized delta rule. There are 1 defect and 0 defect, and learn about 1000 times. If the output of the final output layer is 0.5 or more, there is a flaw, and if it is less than that, there is no flaw.

It is necessary to put a memory in a proper place to calculate the normalized feature amount. As shown in Fig. 5, the memory installation position is A / D converter 2
The output side of 2 and the output side of the differential range calculation units 24 and 26. Range B is for 1 line (1/2, 1/4 lines, etc., but in this case, it is for 1 line), and range A is the differential part of 1 line, so there are multiple lines in memory M _1. For the one line in the calculation units 23 and 24. After calculating, the calculation units 25 and 26 calculate ₂ , x _DR for each range A of the line, and the normalization processing unit 27 calculates (5) above.
Normalized quadratic mean by equation (6) One method is to calculate the differential range X _DRS and input it to the neural network 30.

Using the memory M ₂ , the calculation result for each range A of one line is stored in M ₂ , and when the calculation result for the line, that is, the range B is obtained, the processing unit 27 performs the normalization process.
Is the second method, the calculation results for the range B stored in the memory M ₃ using the memory M _3, of each of the ranges A in the next line calculations and memory M ₃ content (one line before The third method is to perform the normalization processing using the calculation result of the range B).

The above-mentioned third method is an effective method because the memory capacity is small and there is not much difference in the secondary average between the previous line and the current line.

3 and 4 show the false detection rate and the like. Fig. 3 shows the conventional method, the horizontal axis is the threshold value (2,4, ... corresponds to 0.2V, 0.4, ...)
The vertical axis is rate%. If the threshold value shown in the figure is low, the rate of over-detection (marked as defective even though it is not a defect) is high, and if the threshold value is high, the rate of false detection (missing) is high. The threshold for the correct answer rate changes variously depending on the surface properties of the steel sheet and the like.

FIG. 4 is a graph showing the present invention (neural net method) and the conventional method (differential threshold level method) in comparison.
The horizontal axis is the false detection rate, and the vertical axis is the over detection rate. As shown, in the conventional method, when one becomes better, the other becomes worse. On the other hand, in the method of the present invention, as shown in groups G ₁ and G ₂ , both
Can be lowered. It should be noted that G ₂ is a case where the neural network is trained better than G ₁ .

Normalized quadratic mean _2S and normalized differential range
In addition to adopting x _DRS , it is also conceivable to use an appropriate combination including the _third -order average ₃ = E · Σy ³ _i as the feature amount. Next, the result of this combination experiment is shown.

From this table, Case No. 10 shows good results,
The one using the third-order average gives a poor result although it requires a complicated calculation.

In the present invention, not only the flaw signal (the signal in the range A) but also the signals around it (the signal in the range B) are used to determine whether there is a flaw. This makes the threshold unnecessary. The differential range gives height (intensity) information and the quadratic average gives flaw shape information.
Whether or not there is a part that differs from the surroundings,
Is being identified.

In the present invention, by learning the differential waveform whose strength has changed, together with the waveform that has not been changed, even if the input differential waveform is changed in intensity (that is, even if S / N deteriorates). ) A system with little influence on the detection rate can be obtained.
It is quite conceivable that the magnitude of the differential signal will change depending on the installation conditions on the line, the deterioration of the equipment, and the type of the DUT. In the conventional differential threshold method, it is impossible in principle to flexibly respond to these changes,
The detection accuracy deteriorates, and there is a great risk that adjustment (that is, the value of the differential threshold level) will be adjusted after installation.

The differential strength of the data currently used for learning is ± 20
The data that was changed in% was also learned. Then, with respect to this learning data as well, a variation of ± 20% was applied, and the detection rate,
The overdetection rate was evaluated, and it was confirmed that the differential method showed a large fluctuation, while the neural network method showed a very small fluctuation. The following table shows the results.

Table 7 shows the latest data for Tables 2 and 3.

[Effects of the Invention] As described above, according to the present invention, it is possible to perform a steel plate surface flaw inspection capable of suppressing both false detection and over-detection to the minimum without being bothered by threshold setting.

[Brief description of drawings]

FIG. 1 is an explanatory view of a steel plate surface flaw inspection method of the present invention, FIG. 2 is an explanatory view of each portion of the flaw inspection, FIGS. 3 and 4 are graphs showing inspection result examples, and FIG. 5 is a memory installation position. It is a block diagram showing each example of. In FIG. 1, 21 is a differentiating circuit and 22 is an A / D converter.

Claims (3)

[Claims] 1. A steel plate surface is scanned with a laser beam, the reflected light is photoelectrically converted, and A / D converted, and the small range (A) of the steel plate and the large range (B) including a plurality of the ranges are measured. From the converted output, calculate a plurality of feature quantities that represent the reflected light waveform,
The feature quantity is added to the neural network that has been learned in advance and
A steel sheet surface flaw inspection method using a neural network, which is characterized by producing no output. 2. The steel sheet surface flaw inspection method according to claim 1, wherein the feature quantities are a normalized quadratic mean _2S and a normalized differential range x _DRS represented by the following equation. Here, ₂ is the secondary average of A / D conversion output in a small range, Is the quadratic average of the A / D conversion output of the large range B, x _DR is the differential range of the small range A, and σ _XDR is the standard deviation of the differential range of the large range B. 3. A laser oscillator for generating a laser beam for scanning the surface of a steel sheet, a receiver for receiving the laser beam reflected by the surface of the steel sheet, and a circuit for differentiating the output of the receiver and A / D converting the differential output. And means for calculating a secondary average and a differential range for each small range (A) from the A / D converted output, and a large range (B including a plurality of the small ranges from the A / D converted output. ) For calculating the second-order average, the differential range, and the standard deviation, and the normalized second-order average from these second-order averages, and from the small-range (A) differential range and the large-range (B) standard deviation. Normalization processing means for calculating a normalization differential range, and addition of the normalized secondary average and differential range
A steel sheet surface flaw inspection apparatus using a neural network, comprising: a neural network that produces no output.