JP7701208B2 - Automatic assessment system for the deterioration of anchor bolts - Google Patents

Description

This invention relates to an automatic deterioration determination system for anchor bolts, and in particular to an automatic deterioration determination system for anchor bolts that allows an inspector to determine the deterioration state of anchor bolts on-site.

Conventionally, an ultrasonic inspection method for anchor bolts is described in, for example, Japanese Patent No. 6088088 (Patent Document 1). According to this publication, an ultrasonic inspection device for anchor bolts inspects corroded parts of anchor bolts 27 using an ultrasonic probe. The ultrasonic inspection device for anchor bolts includes a cylindrical probe attachment part that can be attached to the head of the anchor bolt and has an inclined surface at the top to which an ultrasonic probe can be attached, a probe rotation jig that rotatably holds the probe attachment part, and a digital ultrasonic flaw detector that receives a reflected echo of an ultrasonic inspection signal from the ultrasonic probe and determines deterioration of the anchor bolt.

Patent No. 6088088

In Patent Document 1, the deterioration of the anchor bolt was determined by receiving the reflected echo of the ultrasonic flaw detection signal from the ultrasonic probe. In such a conventional anchor bolt deterioration determination system, a skilled technician determined the degree of thread damage, etc., based on the inspection results.

As a method for determining the deterioration of anchor bolts, it is possible to create an exploded view of the bolt from the inspection results and illustrate the location and extent of the damage, but there is a problem in that only experienced technicians can determine the condition of the anchor bolt (sound, bolt with initial damage, bolt with damage, dangerous bolt).

This invention was made to solve the problems mentioned above, and aims to provide an automatic deterioration determination system for anchor bolts that allows on-site inspectors to easily determine the deterioration state of anchor bolts.

The automatic deterioration determination system for anchor bolts according to the present invention automatically determines the deterioration level of an anchor bolt based on a pre-created development diagram showing the deterioration level of the anchor bolt. The development diagram is created based on data obtained by detecting the deterioration level of the anchor bolt with an ultrasonic probe, and shows the deterioration level, position, and size of the anchor bolt. The development diagram separates the healthy part of the bolt from the half-depth defective part of the thread and the deep defective part of the thread with a specified color, and the proportion and the occurrence position are read from the development diagram according to the specified color, performing pre-processing, and the deterioration level of the anchor bolt is automatically determined based on the results of the pre-processing.

Preferably, the step of automatically determining the degree of deterioration of the anchor bolt is performed using fuzzy logic.

The degree of deterioration of anchor bolts may be divided into three or more stages, such as healthy and dangerous, or healthy, caution, and dangerous.

The step of determining the degree of deterioration of the anchor bolt using fuzzy logic may be performed in two stages: a first stage and a second stage.

The first stage preferably determines the deterioration of health and danger, and the second stage determines the deterioration of the caution level.

According to this invention, the measurement data is associated with the degree of deterioration based on the judgment of a skilled technician, and if there is measurement data, the degree of deterioration is automatically determined.

As a result, it is possible to provide an automatic system for determining the degree of deterioration of anchor bolts, enabling on-site inspectors to determine the deterioration state of anchor bolts.

1 is a diagram showing an example of a measurement image used in a deterioration determination system according to an embodiment of the present invention; FIG. 1 is a diagram showing an overview of image data used in constructing the system. FIG. 13 is a diagram showing the difficulty of separating white lines; FIG. 1 is a diagram illustrating the concept of expansion processing and contraction processing. FIG. 13 is a diagram for explaining the reason for setting the line interval. FIG. 13 is a diagram showing actual measured values and estimated values of known colors. FIG. 13 is a diagram showing the proportion of each color line for each deterioration determination level. FIG. 1 is a diagram illustrating the concept of fuzzy reasoning. FIG. 1 is a diagram illustrating the concept of applying fuzzy inference. 13 is a detailed flowchart of fuzzy decision making. FIG. 1 is a diagram showing an example of an inspection defect mechanism and an image obtained when measurement is performed in that state. FIG. 13 is a diagram showing a membership function for determining whether or not an inspection is defective. FIG. 13 is a diagram showing a first membership function. FIG. 13 is a diagram showing a second membership function. 4 is a flowchart of a pre-processing of the deterioration determination system according to the embodiment of the present invention. 4 is a flowchart of a fuzzy determination process of the deterioration determination system according to the embodiment of the present invention. FIG. 2 is a diagram showing an overall configuration of an anchor bolt deterioration determination device to which the above-mentioned flowchart is applied.

One embodiment of the present invention will be described below with reference to the drawings. First, the data used in creating this automatic system for determining the degree of deterioration of anchor bolts will be described. The data used this time is as shown in Figure 1, which is a development diagram in which the level of deterioration of the anchor bolt measured by an ultrasonic probe of an anchor bolt deterioration determination device is represented by lines at each position. The horizontal axis represents the angle in the circumferential direction of the anchor bolt, and the vertical axis represents the distance in the depth direction.

The image shown in Figure 1 originally contains blue (healthy area), yellow (Caution Level 1), white (Caution Level 2), and red (dangerous area), but here, each is represented by four lines: thin, dotted, dashed and thick. Here, the thin lines are the beam path of the healthy area echo, the dotted lines indicate areas where the defect has progressed to about half the depth of the thread, the dashed lines indicate areas where the defect has progressed to about the depth of the thread, and the thick lines indicate areas where the defect has progressed beyond the depth of the thread.

Note that in the figure, four lines are used to represent the area: a thin line, a dotted line, a dashed line, and a thick line. However, in reality, the areas indicated by the dashed lines are shown in white as measurement results.

Referring to Figure 1, for example, this is for a screw with a nominal diameter of M22. It shows which of the following data the data corresponds to: healthy, caution level 1, caution level 2, danger, or inspection failure.

Figure 2 shows an overview of the number of pieces of data used in this embodiment from such image data. In Figure 2, a table is displayed indicating whether the measured data for M22, M24, M27, M30, and M36 is sound, caution level 1, caution level 2, dangerous, or inspection failure, and a bar graph of the number of pieces of data shown above the table is displayed. To the right of this is a pie chart showing the totals for each level of bolt deterioration, such as soundness, and a breakdown of each level based on that.

Because the measurement range differed depending on the nominal diameter of the screw, it was necessary to homogenize this data according to a certain rule in order to compile and develop the system. There were a total of 404 images, which were visually sorted by experienced personnel into four levels: healthy level, caution level 1, caution level 2, and danger level, and the results were used as training data. As a result, as shown in Figure 2, approximately 60% were in the healthy level, approximately 25% were in caution level 1, approximately 10% were in caution level 2, and approximately 10% were in the danger level.

Next, we will explain how to count the number of colors in an image. Figure 3 shows the result image (A) and a schematic diagram (B) that shows the intensity of the measured waveform used to create the result image and the colors that correspond to that intensity.

The inventors first considered counting the number of colors in an image. That is, they initially thought that it might be possible to determine deterioration by counting the number of blue, yellow, red, and white lines in an image. However, as shown in Figure 3(B), the white lines are the same color as the background and are difficult to distinguish, so counting the number of white lines became problematic.

First, they considered using information processing to convert just the white lines into a different color, but because they were measuring screws that were spirally machined, the lines displayed in the images obtained from the measurement data were also inclined, and they concluded that it would be difficult to identify the white lines using information processing.

The next suggestion was to change the background color to a different color, but the white color indicates a low echo height, and although there is measurement data for the white parts other than those shown as lines representing the threads, they are white because they have a low echo. Therefore, even if the background color were changed, the white lines would now be the same color as the changed background color, and the same problem would arise.

In addition, because this low echo level may contain noise, it proved difficult to identify and display only the threads as lines from data with low echo levels.

Based on the above, we determined that it would be difficult to compile an exact count of the white areas, and decided to estimate the amount of white lines by estimating the white lines. We also considered two methods for this: an image processing method to estimate the white lines using expansion and contraction, and a method to estimate the white lines using straight line detection to detect thread lines.

Next, we will explain how to estimate the white line using this image processing. Here, we attempted to estimate the white line using image processing techniques such as binarization, expansion processing, and contraction processing, as well as opening and labeling processing that combine these.

Figure 4 shows the concept of expansion and contraction. As shown in Figure 4, in the expansion process, the pixel of interest is replaced with white, and in the contraction process, the pixel of interest is replaced with black. The number of processes must be set to perform the expansion or contraction process, but if it is performed a sufficient number of times to fill in the white areas between the thread lines, then a portion estimated to be a white line will remain. Also, although the thread lines are inclined, the distance between the thread lines is nearly parallel and is nearly constant, so it can be seen that the distance between the thread lines can be calculated.

Next, we will explain the line spacing. By appropriately setting the line spacing (in pixels), the number of times the expansion/contraction process is performed is determined, and this is done to separate the white of the background color from the white of the lines. In other words, the line spacing and the number of times the expansion/contraction process is performed are synonymous. Figure 5 is a diagram that explains the reason for setting the line spacing.

As shown in Figure 5, lines can only be extracted if the expansion/contraction process is performed appropriately (as in the case of the right horizontal arrow in the figure). In contrast, if the expansion/contraction process is performed a number of times greater than the number of times that the next line is reached (as in the case of the right diagonal upward arrow in the figure), the line information will disappear. Conversely, if the expansion/contraction process is performed a number of times that the next line is not reached (as in the case of the right diagonal downward arrow in the figure), the background color will remain as noise. Furthermore, even if a labeling process is performed to surround the line area when the line spacing is not specified appropriately, the line area cannot be accurately obtained.

To verify this estimation method, we applied it to three colors with a fixed number of colors: red, yellow, and blue. The results are shown in Figure 6. In this figure, the horizontal axis shows the number of each color found using this estimation method, and the vertical axis shows the number of each color measured. The correlation coefficient found from this figure was 0.996, indicating an extremely strong positive correlation, and it was found that the estimation of the white line using this method is valid. We also performed estimation using the same process for other colors to ensure the same conditions.

Next, we will explain deterioration judgment based on the estimation of the number of each color. Figure 7 shows the results of indicating the percentage of each image color for each deterioration judgment level, based on the values obtained by estimation through image processing of each color. Note that the size of the judgment area differs depending on the screw diameter, so the results are normalized by the judgment area size.

Referring to Figure 7, the vertical axis shows the proportion of each color.

These results show a tendency for the proportion of colors other than blue to decrease and the proportion of yellow and white to increase as the degree of deterioration progresses, but because the proportions are not divided by deterioration assessment level, it is thought that it would be difficult to determine a threshold value and classify them.

Therefore, we investigated the use of fuzzy inference to determine deterioration. The method of classification using the threshold value shown in the deterioration determination mentioned above is as shown in Figure 8 (A), where children and adults are clearly separated at the threshold age of 20. However, in reality, the boundary between children and adults is vague, with some people considering people to be adults from 18 years old and others considering people to be children up to 22 years old. We considered the idea of fuzzy inference as a method of expressing such ambiguity.

As shown in Figure 8 (B), this fuzzy reasoning is an idea that expresses the previously mentioned thinking of children and adults in a way that allows for both possibilities. The concept that was considered when applying this is shown in Figure 9. As shown in this figure, the horizontal axis shows the conditions under which people judge categories, and the vertical axis shows the degree of agreement with each category.

Next, we will explain the application of fuzzy reasoning. When we asked the experts what they paid attention to in the images when applying fuzzy reasoning, they said that they looked at the amount and size of each of the colors red, yellow, and white in the image. Therefore, we first tried to determine the deterioration level using fuzzy reasoning by applying the following rules to items other than inspection defects.

If the proportion of blue lines, yellow lines, and white lines is high, the inspection is defective.

If the proportion of blue lines is high, the proportion of yellow lines is high, and the proportion of white lines is low, the warning level is 1.

If the proportion of blue lines is large, the proportion of yellow lines is small, and the proportion of white lines is large, the inspection is defective.

If the proportion of blue lines is high, the proportion of yellow lines is low, and the proportion of white lines is low, the warning level is 1.

If the proportion of blue lines is small, the proportion of yellow lines is large, and the proportion of white lines is large, the warning level is 2.

If the proportion of blue lines, yellow lines, and white lines is small, the warning level is 2.

If the ratio of blue lines is small, the ratio of yellow lines is small, and the ratio of white lines is large, it is dangerous.

If the proportion of red lines, yellow lines, and white lines is small, then it is healthy.

Therefore, from a total of 397 pieces of data (226 healthy, 94 caution 1, 39 caution 2, and 38 dangerous), the proportions of healthy, caution 1, caution 2, and dangerous were extracted for each of the blue, yellow, and white data, and a membership function and judgment rules for analysis were set to make a deterioration judgment. The results are shown in Table 1.

The results of judging the training data according to the above rules are shown in Table 2.

In Table 2, the "visual judgment result" is the judgment result by a skilled technician, and the "automatic judgment result" is the inference result according to the above rules. Therefore, referring to Table 2, the accuracy rate is 76.3%, and if a difference of up to one is considered correct, it becomes 99.7%. Here, the accuracy rate is calculated as 225/226 = 99.6%, for example, since there were 225 correct visual judgment results and the automatic judgment results included one item with a warning 1.

The reason for the low accuracy is that while the correct answer rates for healthy and dangerous are close to 100%, Caution 1 and Caution 2 are significantly lower at approximately 40% and 15%, respectively. This makes it necessary to reconsider the membership functions and judgment rules.

In the reexamination, we introduced optimization because it is difficult to set it manually. The optimization targets are the vertices of the membership function and the decision rules.

The vertices of the membership function are 101 levels from 0-100% for blue, and 101 levels from 0 to 10% for yellow and white. The decision rule is four levels based on the state of the consequent.

Therefore, the nine data points at the vertices of the three membership functions for the three input values and the 27 data points in the consequent part of the decision rule were targeted for optimization, and 30 data points randomly extracted from each state, for a total of 120 data points, were used as training data, and the accuracy rate when using each membership function and fuzzy rule was used as the evaluation value.

The results are shown in Table 3.

The overall accuracy rate did not change significantly, but the accuracy of Caution 1 and Caution 2 improved, while the accuracy of the Danger Level dropped significantly.

We built a system aiming to perform four classifications using one fuzzy rule, but to improve the accuracy of the judgment, we decided to perform fuzzy inference again with different parameters for those judged to require caution, and to divide them into caution 1 and caution 2.

A detailed flowchart of the fuzzy judgment for this purpose is shown in Figure 10.

Referring to FIG. 10, in the fuzzy judgment process, fuzzy inference is divided into three stages. First, the first fuzzy inference is performed (S11), where items with poor installation are separated. Then, for items that are not poorly installed, the second fuzzy inference is performed (S12), where items are separated into healthy and dangerous. After that, for items that are determined to require caution, the third fuzzy inference is performed (S13), where they are classified into Caution 1 and Caution 2.

The results are shown in Table 4.

Here, we will explain the case of improper installation where the anchor bolt deterioration judgment device was not installed correctly in the first fuzzy inference. Because improper installation or other factors can cause inaccurate images to be input that make it impossible to judge deterioration, we introduced fuzzy theory to automatically distinguish between such images that are improperly inspected and images that are subject to judgment.

We will now discuss fuzzy inference for determining whether an inspection has been faulty. An inspection fault here refers primarily to a state in which the fixing jig that attaches the sensor (probe) to the anchor bolt is not installed properly (perpendicular to the axis of rotation of the anchor bolt), as shown in Figure 11 (A). The sensor is optimized to scan the threads of the anchor bolt when it is properly installed, so when measurements are made in this state, the ultrasonic waves sent from the sensor may not hit the threads and the reflected echo may be low. Conversely, on the 180-degree opposite side, the ultrasonic waves may hit at a shorter distance than expected, so the reflected echo tends to be high.

When the reflected echo obtained in this way is graphed, it is displayed in white because the reflected echo is small, as shown in Figure 11 (B). If we were to apply the four categories of deterioration judgment without considering the possibility of inspection failure, it would be judged as dangerous, so we created a system that uses fuzzy inference to judge whether or not an inspection failure has occurred.

Experienced technicians often judge whether an inspection is defective based on the white shape, which indicates the smallest reflected echo, and the red shape, which indicates the largest reflected echo, so we constructed the white and red shapes by focusing on their width and height.

The results are shown in Table 5, the rules at that time are shown in Table 6, and the membership functions at that time are shown in Figure 12.

The membership function for the first fuzzy inference is as shown in Figure 13, and the fuzzy rules are as shown in Table 7. The membership function for the second fuzzy inference is as shown in Figure 14, and the fuzzy rules are as shown in Table 8.

Figure 15 shows a pre-processing flowchart that was created based on the above points. Referring to Figure 15, first, a development diagram in which the deterioration level of the anchor bolts shown in Figure 1 is represented by lines for each position is input, and based on this, a waveform portion is extracted (S21) as shown in Figure 3 (B), and as pre-processing, the lines of each color are emphasized using an image processing technique (S22-S23). This is performed a set number of times. After that, opening processing (S25), labeling processing (S26), and processing to obtain information such as the color of the lines of interest (S27) are performed, and information such as area is extracted from the emphasized data, and this is performed for all colors.

The opening process refers to a process that performs contraction and expansion once each to remove noise, and the labeling process refers to a process that identifies the position and area of each block.

Here, the data obtained by total color tallying is, for example, blue area ratio 99.971, yellow area 0.066, white area 0.029, and red area 0.00.

Next, the fuzzy judgment process will be described. FIG. 16 is a detailed flowchart for explaining the process in the fuzzy judgment process shown in FIG. 10. Referring to FIG. 16, in the fuzzy judgment process, first, the data obtained in the preprocessing, such as the blue area ratio of 99.971, the yellow area of 0.066, the white area of 0.029, and the red area of 0.00, are input, and the first stage of fuzzy inference (S30) is performed to judge whether the inspection is defective or the measurement target (S31). If the measurement target is ("measurement target" in S31), the second stage of fuzzy inference (S32) is performed, and the second stage of deterioration judgment processing is performed (S33). Here, it is judged whether it is healthy, dangerous, or requires caution. If it is judged to be healthy or dangerous, the process ends.

Only those judged to require caution go to the third stage of fuzzy inference (S34), where the second stage of deterioration judgment processing is performed (S35), and the item is judged to require either caution 1 or caution 2.

Next, an ultrasonic inspection device to which this system is applied will be described. Figure 17 is a diagram showing the overall configuration of an anchor bolt deterioration determination device 10 to which the above-mentioned flowchart is applied. Referring to Figure 17, the anchor bolt deterioration determination device 10 includes a control unit 20 including a CPU, a detection unit 21 having an ultrasonic probe connected to the control unit 20, and an application 22 included in the control unit 20, and the above-mentioned automatic determination system for the deterioration degree of the anchor bolt is included in this application. In addition, this application performs processes from creating a development diagram showing the degree of deterioration of the anchor bolt to determining the deterioration of the anchor bolt, and outputs the results.

Although the embodiment of the present invention has been described with reference to the drawings, the present invention is not limited to the illustrated embodiment. Various modifications can be made to the illustrated embodiment within the same scope as the present invention or within an equivalent scope.

According to this invention, the anchor bolt deterioration determination device allows workers to determine the degree of deterioration on-site, and is therefore advantageously used as an automatic determination system for the degree of deterioration of anchor bolts.

10 Anchor bolt deterioration determination device, 20 Control unit, 21 Detection unit, 22 Application.

Claims (5)

An automatic determination system for a deterioration degree of an anchor bolt, which automatically determines a deterioration degree of an anchor bolt based on a development diagram showing a deterioration degree of the anchor bolt that is created in advance,
The development diagram is created based on data obtained by detecting the degree of deterioration of the anchor bolts using an ultrasonic probe, and indicates the degree of deterioration of the anchor bolts, their positions, and their sizes,
In the developed view, a healthy portion of the bolt, a half-depth missing portion of the thread, and a depth missing portion of the thread are separated by a predetermined color,
A pre-processing step is performed to read the ratio and the occurrence position of the predetermined color from the development view according to the predetermined color;
An automatic deterioration determination system for anchor bolts that automatically determines the deterioration level of anchor bolts based on the results of pre-processing. The system for automatically determining the deterioration level of an anchor bolt according to claim 1, wherein the step of automatically determining the deterioration level of the anchor bolt is performed using fuzzy theory. The system for automatically determining the deterioration level of an anchor bolt according to claim 2, wherein the deterioration level of an anchor bolt is divided into three stages: healthy, caution, and danger. The system for automatically determining the deterioration level of an anchor bolt according to claim 3, wherein the step of determining the deterioration level of the anchor bolt using fuzzy logic is performed in a first stage and a second stage. 5. The system for automatically determining a deterioration degree of an anchor bolt according to claim 4, wherein the first stage determines whether the deterioration is sound or dangerous, and the second stage determines whether the deterioration is at a caution level.

Images