Regular model-based nuclear rod end plug welding seam ray digital imaging defect identification method
By introducing a rule-based model method in the detection of weld defects of the end plug of nuclear fuel rod, combined with the target detection algorithm and expert rules, the problem of subjective judgment and low accuracy in the detection is solved, and efficient and reliable defect identification is achieved.
Patent Information
- Application Number
- CN202411748532.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-02
- Publication Date
- 2025-03-25
AI Technical Summary
The prior art has subjective judgment influence, false alarms and missed reports in the detection of end plug weld defects of nuclear fuel rods. Moreover, due to the few defect images, the labeling is difficult, resulting in low detection accuracy.
The method based on the rule model is adopted, combined with the object detection algorithm and expert rules, defect recognition is achieved by obtaining single-stick images, detecting the location of defect samples, and further identifying defect types based on the rules characteristics.
It improves the accurate identification ability of weld defects, reduces the impact of subjective judgments, enhances the reliability of detection results, and achieves rapid execution and image defect positioning on smaller-scale hardware facilities.
Smart Images
Figure CN119693315A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of nuclear power technology, and in particular relates to a method for identifying defects in nuclear fuel rod end plug welds by radiographic digital imaging based on a rule model. Background Art
[0002] Due to different environmental conditions and welding technology differences, welding defects are inevitable during the welding process of nuclear fuel rod end plugs. Therefore, it is crucial to ensure the quality of the weld to maintain the reliability and safety of the structure. X-ray inspection is one of the main methods for welding quality assessment, and researchers have made significant progress in automatically detecting weld defects.
[0003] In the field of nuclear fuel rod end plug weld defect detection, traditional methods usually require experienced inspectors to make subjective judgments, which makes the detection results easily affected by individual subjective factors. In addition, a lot of marking work is required, and it is easily affected by factors such as the on-site detection environment, which leads to problems such as false positives and missed positives. In order to improve the accuracy of weld quality, researchers introduced models such as convolutional neural networks (CNN) to achieve intelligent detection.
[0004] At present, the target detection algorithms based on deep learning include single-stage and dual-stage modes. Among them, SSD and YOLO algorithms are widely used and rapidly developing single-stage target detection algorithms. Although they have fewer parameters and faster training and detection speeds, their detection accuracy is relatively low. Although dual-stage algorithms such as Faster-RCNN have slightly higher detection accuracy, they are slower. In addition, due to the small number of nuclear fuel rod end plug defect images in the industrial production process and the difficulty of labeling, a variety of data enhancement methods are generally used in the image preprocessing process to expand the defect database, but this will affect the location information of the defect data and interfere with the algorithm model. These technologies have not yet been applied to the automatic recognition of the X-ray digital imaging detection results of the fuel rod end plug weld. Summary of the invention
[0005] In order to overcome the problems existing in the related art, a method for identifying defects in nuclear fuel rod end plug weld radiographic digital imaging based on a rule model is provided.
[0006] According to one aspect of an embodiment of the present disclosure, a method for identifying defects in a nuclear fuel rod end plug weld by radiographic digital imaging based on a rule model is provided, the method comprising:
[0007] Step 1, obtaining a single-stick image;
[0008] Step 2: Use the target detection algorithm network to obtain the defect sample location;
[0009] Step 3: Use the rule model to further identify the defect type, that is, generate the rule features based on the actual fuel rod defect sample characteristics after being refined by artificial expert experience.
[0010] In a possible implementation, the types of defects that may occur in the girth weld area include pores, gas expansion, incomplete penetration and tungsten inclusions; three images are collected for each nuclear fuel rod;
[0011] The defect type is pores. The defect characteristics are grayscale dots appearing in the evaluation area. The vertical heights of the dots on the three images are relatively consistent, but the horizontal positions change after rotation. The judgment rule is to combine the three images for comprehensive evaluation and the dots must appear at least twice.
[0012] The defect type is bulging. The defect characteristics are spots with increased grayscale near the cladding side within 2 mm below the center line of the weld. Also, within 2 mm below the center line of the weld, a band-like image with increased horizontal grayscale appears. The judgment rule is that if it appears once in 3 images, it is judged as bulging.
[0013] The defect type is incomplete penetration. The defect features are thin lines with increased grayscale appearing on the center line of the weld; black spots appearing at the intersection of the center line of the weld and the extension line of the inner wall of the cladding; and vertical lines appearing on the center line of the weld, located near the junction of the cladding and the end plug. The judgment rule is that if it appears once in 3 images, it is judged as incomplete penetration.
[0014] The defect type is tungsten inclusion. The defect feature is white spots with reduced grayscale appearing in the evaluation area. The judgment rule is to conduct a comprehensive evaluation based on 3 images, which must appear at least 2 times.
[0015] In a possible implementation, the defect priority determination in the rule model includes:
[0016] If there is a black spot at the intersection of the center line and the extension line of the inner wall of the cladding, it is judged as incomplete penetration; if there is a black spot at other positions on the center line, it is judged as a pore;
[0017] If black spots appear on the extension of the inner wall of the cladding within 2 mm below the center line, it is first judged as bulging. If black spots appear again at the same high position in other images, it is judged as pores.
[0018] In a possible implementation, the defect feature quantity includes the defect center position, aspect ratio, defect grayscale feature / grayscale difference with the background, and defect area size.
[0019] In one possible implementation, the defect center position includes the relative height characteristics to the weld center line and the distance to the cladding edge; the defect center position Determine using the following formula:
[0020]
[0021] Among them, x min is the minimum value of the horizontal axis, x max is the maximum value of the horizontal axis, y min is the minimum value of the ordinate, y max is the maximum value of the vertical coordinate;
[0022] After determining the defect center position, determine whether the defect center position is on the weld center line or the extension line of the cladding on both sides.
[0023] In one possible implementation, the aspect ratio is determined using the following formula:
[0024]
[0025] Where ε is the equivalent area aspect ratio of the defect, L is the defect length, and W is the defect width;
[0026] When the aspect ratio ε>=K, the defect is a strip defect, and when the aspect ratio ε<K, the defect is a circular defect.
[0027] In a possible implementation, the grayscale feature of the defect / the grayscale difference with the background is determined by the following formula:
[0028]
[0029] Among them, e is the grayscale difference between the defect and the background, A is the average grayscale value of all pixels in the defect area, and B is the average grayscale value of all pixels in the background area.
[0030] In one possible implementation, the defect area is determined using the following formula:
[0031] S=w×h
[0032] Wherein, w is the width of the defect circumscribed rectangle, and h is the length of the rectangle, both in pixels.
[0033] The beneficial effects of the present disclosure are:
[0034] 1. The present invention mainly solves the problem that defects in the nuclear fuel rod end plug weld image are difficult to classify or the classification accuracy is not high by introducing expert rules and using rule model algorithms;
[0035] 2. The present invention is based on a general single-stage deep learning target detection algorithm, which solves the problems of small target area of weld defects and low contrast between defects and background, and can realize rapid positioning and identification of defect positions. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1The present invention is a flowchart of a method for identifying defects of nuclear fuel rod end plug weld by radiographic digital imaging based on a rule model according to an exemplary embodiment.
[0037] Figure 2 It is a schematic diagram of the center line of the end plug weld and the extension line of the cladding on both sides.
[0038] Figure 3 It is the defect rule tree model.
[0039] Figure 4 Schematic diagram of typical weld defects. DETAILED DESCRIPTION
[0040] The present disclosure is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0041] Figure 1 is a flow chart of a method for identifying defects in a nuclear fuel rod end plug weld using digital radiographic imaging based on a rule model according to an exemplary embodiment. Figure 1 As shown, the method includes:
[0042] Step 1, obtain a single rod image. In order to better help the inspection personnel to conduct qualitative analysis of the defect type, the end plug weld is generally subjected to double-wall double-image transillumination on the production line, and 120°×3 transilluminations are performed. Therefore, the present disclosure collects 3 images for each single rod.
[0043] Step 2: Obtain the defect sample location through the target detection algorithm network (such as faster-rcnn, ssd, yolo series, this disclosure uses the yolox target detection algorithm).
[0044] Step 3: The defect types are further identified through the rule model, i.e., the rule features generated based on the actual fuel rod defect sample characteristics after being refined by human expert experience.
[0045] Types of defects that may occur in the girth weld area include porosity, gas expansion, lack of penetration and tungsten inclusions.
[0046] The rule model is shown in the following table.
[0047]
[0048] Defect priority determination in the rule model includes.
[0049] If there is a black spot at the intersection of the center line and the extension line of the inner wall of the cladding, it is judged as incomplete penetration; if a black spot appears at other positions on the center line, it is judged as a porosity.
[0050] If black spots appear on the extension of the inner wall of the cladding within 2 mm below the center line, it is first judged as bulging. If black spots appear again at the same high position in other images, it is judged as pores.
[0051] Based on the above defect characteristics and combined with expert experience, the defect features that can be extracted include the defect center position, aspect ratio, defect grayscale characteristics / grayscale difference with the background, and defect area size.
[0052] The center position of the defect, including the relative height characteristics to the weld centerline, the distance to the cladding edge, etc.;
[0053] Defect center location Determined using the following formula.
[0054]
[0055] Among them, x min is the minimum value of the horizontal axis, x max is the maximum value of the horizontal axis, y min is the minimum value of the ordinate, y max is the maximum value of the vertical axis.
[0056] like Figure 2 As shown, after determining the center position of the defect, it is determined whether the center position of the defect is on the center line of the weld or on the extension line of the cladding on both sides.
[0057] The aspect ratio is determined using the following formula.
[0058]
[0059] Among them, ε is the equivalent area aspect ratio of the defect, L is the defect length, and W is the defect width.
[0060] When the aspect ratio ε>=K, the defect is a strip defect, and when the aspect ratio ε<K, the defect is a circular defect.
[0061] The grayscale feature of the defect / the grayscale difference with the background is determined by the following formula. It is used to judge the tungsten inclusion / non-tungsten inclusion features.
[0062]
[0063] Among them, e is the grayscale difference between the defect and the background, A is the average grayscale value of all pixels in the defect area, and B is the average grayscale value of all pixels in the background area.
[0064] The defect area is determined using the following formula.
[0065] S=w×h
[0066] Wherein, w is the width of the defect circumscribed rectangle, and h is the length of the rectangle, both in pixels.
[0067] Finally, the three images are combined at the same time, and the samples are classified through an integrated tree model (such as random forest and GBDT) to obtain the final defect classification information.
[0068] The present disclosure introduces expert rules to implement a defect recognition algorithm based on a rule model, which can effectively solve the situation in which the defect type is inaccurately identified in an industrial environment due to the small scale of image data and the loss of defect location information after the introduction of data enhancement and expansion of the data model. At the same time, different expert rules can be formulated according to the characteristics of different fuel rod images, and then the algorithm universality can be adjusted through external model parameters and other methods; the present disclosure introduces a general deep learning single-stage target detection algorithm, such as YOLOX, SSD, etc., which can be quickly executed on smaller-scale hardware facilities and image defect positioning. The detection speed is increased by more than 50% compared to the two-stage model.
[0069] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.
Claims
1. A method for identifying defects in nuclear fuel rod end plug welds by radiographic digital imaging based on a rule model, characterized in that: The method comprises: Step 1, obtaining a single-stick image; Step 2: Use the target detection algorithm network to obtain the defect sample location; Step 3: Use the rule model to further identify the defect type, that is, generate the rule features based on the actual fuel rod defect sample characteristics after being refined by artificial expert experience.
2. The method according to claim 1, characterized in that Possible defects in the girth weld area include pores, bulging, incomplete penetration, and tungsten inclusions. Three images are collected for each nuclear fuel rod. The defect type is pores. The defect characteristics are grayscale dots appearing in the evaluation area. The vertical heights of the dots on the three images are relatively consistent, but the horizontal positions change after rotation. The judgment rule is to combine the three images for comprehensive evaluation and the dots must appear at least twice. The defect type is bulging. The defect characteristics are spots with increased grayscale near the cladding side within 2 mm below the center line of the weld. Also, within 2 mm below the center line of the weld, a band-like image with increased horizontal grayscale appears. The judgment rule is that if it appears once in 3 images, it is judged as bulging. The defect type is incomplete penetration. The defect features are thin lines with increased grayscale appearing on the center line of the weld; black spots appearing at the intersection of the center line of the weld and the extension line of the inner wall of the cladding; and vertical lines appearing on the center line of the weld, located near the junction of the cladding and the end plug. The judgment rule is that if it appears once in 3 images, it is judged as incomplete penetration. The defect type is tungsten inclusion. The defect feature is white spots with reduced grayscale appearing in the evaluation area. The judgment rule is to conduct a comprehensive evaluation based on 3 images, which must appear at least 2 times.
3. The method according to claim 1, characterized in that Defect priority determination in the rule model includes: If there is a black spot at the intersection of the center line and the extension line of the inner wall of the cladding, it is judged as incomplete penetration; if there is a black spot at other positions on the center line, it is judged as a pore; If black spots appear on the extension of the inner wall of the cladding within 2 mm below the center line, it is first judged as bulging. If black spots appear again at the same high position in other images, it is judged as pores.
4. The method according to claim 1, characterized in that: The defect characteristics include the defect center position, aspect ratio, defect grayscale characteristics / grayscale difference with the background, and defect area size.
5. The method according to claim 4, characterized in that The center position of the defect, including the relative height characteristics to the weld centerline and the distance to the cladding edge; The center position of the defect Determine using the following formula: Among them, x min is the minimum value of the horizontal axis, x max is the maximum value of the horizontal axis, y min is the minimum value of the ordinate, y max is the maximum value of the vertical coordinate; After determining the defect center position, determine whether the defect center position is on the weld center line or the extension line of the cladding on both sides.
6. The method according to claim 4, characterized in that The aspect ratio is determined using the following formula: Where ε is the equivalent area aspect ratio of the defect, L is the defect length, and W is the defect width; When the aspect ratio ε>=K, the defect is a strip defect, and when the aspect ratio ε<K, the defect is a circular defect.
7. The method according to claim 4, characterized in that The grayscale characteristics of the defect / the grayscale difference with the background is determined by the following formula: Among them, e is the grayscale difference between the defect and the background, A is the average grayscale value of all pixels in the defect area, and B is the average grayscale value of all pixels in the background area.
8. The method according to claim 4, characterized in that The defect area is determined using the following formula: S=w×h Wherein, w is the width of the defect circumscribed rectangle, and h is the length of the rectangle, both in pixels.