JP2021139821A - Ultrasonic flaw detection method for round bar material - Google Patents

Abstract

To provide an ultrasonic flaw detection method for a round bar material enabling a surface flaw and a surface layer flaw to be promptly and surely discriminated.SOLUTION: A lateral wave ultrasonic beam Ub is scanned in a direction along a peripheral surface of a round bar material M while transmitting the lateral wave ultrasonic beam Ub to a round bar material M.A reflected ultrasonic wave reflected and returned by a surface layer flaw M1 of the round bar material M or a surface layer flaw M2 generated within the round bar material M immediately below a surface is received and denoted as a flaw echo signal. A flaw echo image is created by using, as axes of a two-dimensional image, each scanning position of the oblique angle flaw detection ultrasonic wave (Ub) and a time from transmitting the lateral wave ultrasonic beam Ub at the scanning position until obtaining the flaw echo signal. A required number of flaw echo images are given to a neural network 3 as learning data for learning, and a new flaw echo image is given to the neural network 3 which has learnt the learning data, to discriminate whether a flaw corresponding to the flaw echo image is a surface flaw M1 or a surface layer flaw M2.SELECTED DRAWING: Figure 1

Description

The present invention relates to an ultrasonic flaw detection method for a round bar material, and more particularly to an ultrasonic flaw detection method capable of satisfactorily distinguishing and detecting a surface defect and a surface layer defect generated inside the round bar material immediately below the surface.

When detecting a flaw near the surface of the round bar, a transverse wave of the ultrasonic beam Ub for flaw detection is used and the refraction angle (sector scan angle) is set to about 45 degrees as shown in FIG. However, in this method, a surface defect M1 (FIG. 7 (1)) that opens on the surface of the round bar M and a surface defect M2 (FIG. 7 (2)) that occurs immediately below the surface of the round bar M are formed. Since echo signals (FIGS. 8 (1) and 8 (2)) may appear at almost the same size and at the same time zone, it is often difficult to distinguish between the two.

Therefore, for example, in Patent Document 1, different sector scan angles are set, and when the magnitude of the defect echo signal obtained at each sector scan angle exceeds a predetermined threshold value, it is determined that there is a surface defect or a surface defect, respectively. The method of flaw detection is disclosed.

JP 2012-225887

However, in the above-mentioned conventional method, since it is necessary to change the sector scan angle and repeat the same flaw detection, it takes time to detect the flaw, and there is a problem that the round bar material flowing through the line cannot be quickly detected. ..

Therefore, the present invention solves such a problem, and an object of the present invention is to provide an ultrasonic flaw detection method for a round bar material capable of quickly and surely discriminating between surface defects and surface surface defects.

In order to achieve the above object, in the first invention, while transmitting the oblique angle flaw detection ultrasonic wave (Ub) toward the round bar member (M), this is along the peripheral surface of the round bar member (M). Scanned in the direction, the surface defect (M1) of the round bar material (M) or the surface layer defect (M2) generated inside the round bar material (M) just below the surface receives the reflected ultrasonic wave reflected and returned to the defect. As an echo signal, a flaw echo image is created from each scanning position of the oblique angle flaw detection ultrasonic wave (Ub) and the time change of the signal intensity of the flaw echo signal at the scanning position, and a required number of the flaw echo images are created. Is given to the neural network (3) as training data for training, and a new flaw echo image is given to the trained neural network (3), and the flaw corresponding to the flaw echo image is the surface flaw (M1). ) Or surface defect (M2).

In the first invention, a flaw echo image is created, and a new flaw echo image is given to the neural network trained in advance by the flaw echo image to determine whether the flaw corresponding to the flaw image is a surface defect or a surface defect. Since the discrimination is made, the surface defect and the surface layer defect can be discriminated quickly and surely.

In the second invention, the phased array probe (1) is used for transmitting and receiving the ultrasonic wave for oblique angle flaw detection.

According to the second invention, the phased array probe can transmit and scan the ultrasonic wave for oblique angle flaw detection, and receive the reflected ultrasonic wave easily and compactly.

In the third invention, the number of flaw echo signals extracted to obtain the flaw echo image can be changed in magnitude according to the purpose of use.

According to the third invention, by appropriately changing the number of the extracted defect echo signals, the accuracy of discrimination and the speed of discrimination, which is in a trade-off relationship with the discrimination speed, can be appropriately selected according to the flaw detection target. Can be done.

In the fourth invention, in the flaw echo image, a portion having a high signal intensity of the flaw echo signal is displayed with high brightness, and a portion having a low signal intensity of the flaw echo signal is displayed with low brightness.

In the fifth aspect of the present invention, in the flaw echo image, a portion having a high signal intensity of the flaw echo signal is displayed with a low brightness, and a portion having a low signal intensity of the flaw echo signal is displayed with a high brightness.

In the sixth invention, the round bar material (M) is a black skin material.

The reference numerals in parentheses indicate the correspondence with the specific means described in the embodiments described later for reference.

As described above, according to the ultrasonic flaw detection method for round bar members of the present invention, surface defects and surface surface defects can be quickly and reliably discriminated. More specifically, a round bar material that has undergone a rolling step, a heat treatment step, and a bending straightening step and whose surface is covered with black rust, which is a metal oxide, can be targeted. Since the black skin material has black rust on the outermost surface layer, it is particularly difficult to distinguish between surface defects and surface layer defects. However, by applying the method of the present invention, surface defects and surface layer defects can be quickly and reliably discriminated. It becomes possible to do.

It is a figure which shows the structure of the apparatus which carries out the method of this invention. It is a figure which shows an example of a defect echo image. It is a figure which shows the image which cut out the predetermined region containing the flaw echo signal from the flaw echo image. It is a figure which shows the schematic structure of the convolutional neural network. It is a figure which shows another example of a defect echo image. It is a figure which shows still another example of a defect echo image. It is a schematic cross-sectional view which shows the conventional flaw detection method. It is a figure which shows the flaw echo signal in the conventional flaw detection method.

The embodiments described below are merely examples, and various design improvements made by those skilled in the art within the scope of the present invention are also included in the scope of the present invention.

FIG. 1 shows the configuration of an apparatus for carrying out the ultrasonic flaw detection method of the present invention. In FIG. 1, a phased array probe 1 is provided so as to face the outer peripheral surface of the metal round bar M. In the phased array probe 1, a large number of ultrasonic vibrators (not shown) are adjacent to each other in a known structure so as to form a transmission / reception surface 1a that curves according to the outer circumference of the round bar member M, and are adjacent to each other. By vibrating a predetermined number of ultrasonic vibrators with an excitation signal having a predetermined time difference output from a control device 2 having a built-in computer, a round bar having a sector scan angle of approximately 45 degrees in the present embodiment. A transverse ultrasonic beam Ub, which is an ultrasonic for oblique flaw detection, is generated that converges in a region including the surface of the material and the surface layer immediately below the material. Then, by sequentially moving the range of a predetermined number of ultrasonic vibrators to be vibrated, the transverse ultrasonic beam Ub is scanned in an angle range D of about 90 degrees, and the surface flaws M1 and the surface flaws M2 in this range are scanned. The reflected ultrasonic wave is received again by the phased array probe 1 and output to the control device 2 as a defect echo signal.

When detecting surface defects M1 and surface defects M2 on the entire circumference of the round bar material M, a plurality (4) of phased array probes 1 having the same configuration are provided along the entire circumference of the round bar material M. It is provided or the round bar M is rotated, or the phased array probe 1 is swiveled around the round bar M.

FIG. 2 shows a flaw echo image (B scope) drawn in the control device 2 with the scanning position as the horizontal axis and the time change of the signal intensity of the flaw echo signal as the vertical axis by scanning the ultrasonic beam Ub. .. Here, the white line portion of the region surrounded by the broken line in FIG. 2 is the portion corresponding to the flaw echo signal. The defect echo image in this case was obtained at a scanning position interval of 97 lines (scanning position) / 45 degrees.

Here, in the defect echo signal shown in FIGS. 8 (1) and 8 (2) described above, since the waveform of the reflection intensity is converted by an absolute value, each of the positive waveform and the negative waveform is not represented. Since the resolution is low, the accuracy of defect determination is affected. On the other hand, in the flaw echo image (B scope) shown in FIG. 2, the absolute value is not converted, and both the positive flaw echo signal and the negative flaw echo signal are reflected in the signal strength. From this, it is clear that the resolution is doubled as compared with the defect echo signal of FIGS. 8 (1) and 8 (2), and the resolution is high.

Further, in the flaw echo image, the portion where the signal strength of the flaw echo signal is high is displayed with high brightness, and the portion where the signal strength of the flaw echo signal is low is displayed at low brightness, so that high resolution can be obtained. It can be suitably used as a data image. In the flaw echo image, even if the portion where the signal strength of the flaw echo signal is high is displayed at low brightness and the portion where the signal strength of the flaw echo signal is low is displayed at high brightness, the same high resolution can be obtained. It can be suitably used as a training data image.

A predetermined region X (FIG. 2) containing a defect echo signal of a predetermined number of obtained defect echo images is cut out for surface defects M1 and surface defect M2 as shown in FIGS. 3 (1) and 3 (2), respectively, and appropriate for each. Divide into a number of training data images and test data images. As an example, 309 learning data images of the surface defect M1 and 1236 test data images were acquired, 260 learning data images of the surface defect M2, and 1040 test data images were acquired.

Then, a training data image (309) is generated on a convolutional neural network (CNN) having a known configuration having an appropriate number of convolutional layers 31, a pooling layer 32, and a fully connected layer 33, as shown in FIG. 4 constructed in the control device 2. After the training was given by giving (sheet + 260 images), a test data image (1236 images + 1040 images) was given to obtain a test correct answer rate. This is shown in Table 1 (1).

As shown in Table 1 (1), the test accuracy rate for both the surface defect M1 and the surface layer defect M2 is 100%, and very good judgment results are obtained. However, on the other hand, the data transfer speed, which is a problem when considering the use on a production line, is relatively slow, and the processing time for learning model generation is also relatively long. The data transfer speed ratio and the processing time ratio for learning model generation in Table 1 are the ratios when the one in Table 1 (3) described later is set to 1.0.

Therefore, the predetermined region X (FIG. 2) including the defect echo signal is extracted only at the scanning positions of the three lines to obtain a defect echo image (FIG. 5), and 372 learning data images of the surface defect M1 are obtained from this. , 1125 test data images were acquired, 299 learning data images of surface defect M2 were acquired, and 900 test data images were acquired. Subsequently, CNN was given each of the above training data images (372 + 299) for training, and then a test data image (1125 + 900) was given to obtain a test accuracy rate. This is shown in Table 1 (2). According to this, the test correct answer rate of the surface defect M1 was 100%, and the test correct answer rate of the surface defect M2 was 99.4%. Both the data transfer rate and the processing time for learning model generation have been greatly improved.

Further, the predetermined region X (FIG. 2) including the flaw echo signal is extracted only at the scanning position of one line to obtain a flaw echo image (FIG. 6), and 298 learning data images of the surface flaw M1 and test data are obtained. 1203 images were acquired, 246 learning data images of the surface defect M2, and 992 test data images were acquired. Subsequently, CNN was given each of the above training data images (298 + 246) for training, and then a test data image (1203 + 992) was given to obtain a test accuracy rate. This is shown in Table 1 (3). According to this, the test accuracy rate of the surface defect M1 was 98.3%, and the test accuracy rate of the surface defect M2 was 98.0%. Both the data transfer rate and the processing time for training model generation have been further improved.

In this way, increasing the number of flaw echo signals extracted to obtain the flaw echo image given to the CNN (increasing the definition of the flaw echo image) improves the CNN test accuracy rate (judgment result), but on the other hand. Since the data transfer speed is slow and the processing time for generating the training model is long, it is preferable to optimally select the number of flaw echo signals to be extracted in order to obtain the flaw echo image to be given to the CCN according to the application.

1 ... Phased array probe, 2 ... Control device, 3 ... Convolutional neural network, M ... Round bar material, M1 ... Surface flaw, M2 ... Surface flaw, Ub ... Transverse wave ultrasonic beam (ultrasonic wave for oblique flaw detection).

Claims (6)

While transmitting an oblique angle flaw detection ultrasonic wave toward the round bar material, it is scanned in a direction along the peripheral surface of the round bar material, and it occurs in a surface defect of the round bar material or inside the round bar material just below the surface. The reflected ultrasonic waves reflected and returned by the surface flaws are received and used as a flaw echo signal. Is created, the required number of the flaw echo images are given to the neural network as training data for training, and a new flaw echo image is given to the trained neural network to correspond to the flaw echo image. An ultrasonic flaw detection method for a round bar material, characterized in that a surface defect or a surface defect is discriminated. The ultrasonic flaw detection method for a round bar material according to claim 1, wherein a phased array probe is used for transmitting and receiving the ultrasonic flaw detection for bevel angle. The ultrasonic flaw detection method for a round bar material according to claim 1 or 2, wherein the number of flaw echo signals extracted to obtain the flaw echo image can be changed in magnitude according to the purpose of use. The flaw echo image according to any one of claims 1 to 3, wherein a portion having a high signal strength of the flaw echo signal is displayed with high brightness, and a portion having a low signal strength of the flaw echo signal is displayed with a low brightness. Ultrasonic flaw detection method for round bar materials. The flaw echo image according to any one of claims 1 to 3, wherein a portion having a high signal intensity of the flaw echo signal is displayed with a low brightness, and a portion having a low signal strength of the flaw echo signal is displayed with a high brightness. Ultrasonic flaw detection method for round bar materials. The ultrasonic flaw detection method for a round bar according to any one of claims 1 to 5, wherein the round bar is a black skin material.

Images