Steel plate flaw detection method
Through the two-stage flaw detection device combined with high-definition camera and ultrasonic probe, combined with image and ultrasonic map analysis, the error problem of existing ultrasonic flaw detection methods when detecting thicker steel plates is solved, achieving high-accurate flaw detection of steel plates and improving production efficiency.
Patent Information
- Application Number
- CN202410929831.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-11
- Publication Date
- 2024-11-08
AI Technical Summary
The existing ultrasonic flaw detection methods have errors when detecting thicker steel plates, resulting in low detection accuracy and inability to effectively detect internal and surface defects of the steel plates.
Using the two-stage flaw detection device method, firstly, the images on the steel plate surface are collected through a high-definition camera and image comparison and analysis are performed. Then, ultrasonic probe is used for detection, combined with the analysis of images and ultrasonic maps, the defect types and locations of the steel plate are judged through the convolutional neural network. Ensure the accuracy of the test results.
It improves the detection accuracy of steel plate flaw detection, ensures accurate detection of surface and internal defects, avoids misjudgment, and significantly improves production efficiency.
Abstract
Description
Technical Field
[0001] The invention belongs to the field of steel plate production, in particular to a steel plate flaw detection method. Background Art
[0002] Steel plates are currently necessary materials for production in all walks of life, especially special steel plates, which have very high performance requirements and are required to be free of defects on the surface and inside of the steel plates. Therefore, flaw detection inside and outside the steel plates is a necessary operation. Currently, the more common flaw detection methods include ultrasonic flaw detection, magnetic particle flaw detection, eddy current flaw detection, etc. At present, ultrasonic flaw detection is the most widely used because it can detect thicker steel plates. However, errors may occur during the ultrasonic flaw detection process. Abnormal ultrasonic patterns may appear, but actual sampling and detection finds no defects, resulting in low detection accuracy. Summary of the invention
[0003] The invention provides a steel plate flaw detection method to solve the defects in the prior art.
[0004] The present invention is achieved through the following technical solutions:
[0005] A steel plate flaw detection method comprises the following steps:
[0006] Step 1: Clean the surface of the steel plate to be tested to ensure that there is no foreign matter attached to the surface of the steel plate;
[0007] Step 2: The steel plate to be inspected after being cleaned in step 1 is transported to the first steel plate flaw detection device through a transport roller. The first steel plate flaw detection device is divided into several areas. Each area of the first steel plate flaw detection device is provided with a high-definition camera, and the high-definition camera collects images of the steel plate to be inspected;
[0008] Step 3: After the first steel plate flaw detection device in step 2 completes image acquisition, it is sent to the second steel plate flaw detection device through a transport roller. The second steel plate flaw detection device is also divided into several areas, and the number and area of the areas are the same as those of the first steel plate flaw detection device. An ultrasonic probe is set in each area of the second steel plate flaw detection device to perform ultrasonic flaw detection on the steel plate to be detected;
[0009] Step 4: Compare and analyze the images collected by the first steel plate flaw detection device in each area with the images in the steel plate defect image set, and send them to the second steel plate flaw detection device for detection after the detection results are obtained. The ultrasonic spectrum obtained by the ultrasonic probe in each area of the second steel plate flaw detection device is analyzed, and the defect characteristic peaks of the abnormal ultrasonic spectrum and the steel plate defect ultrasonic spectrum in the steel plate defect ultrasonic spectrum set are analyzed and compared to determine the type of steel plate defect, and then the image analysis results of each area and the ultrasonic spectrum analysis results of each area are combined to complete the judgment of steel plate flaw detection.
[0010] In the steel plate flaw detection method as described above, the areas divided in the first steel plate flaw detection device and the second steel plate flaw detection device in step two and step three correspond to the same steel plate positions to be detected, the number of areas divided in the first steel plate flaw detection device and the second steel plate flaw detection device is 9-36, and the length range of the areas divided in the first steel plate flaw detection device and the second steel plate flaw detection device is 60-100 cm, and the width range is 60-100 cm.
[0011] In the steel plate flaw detection method as described above, in the step 2, the high-definition camera in each area takes at least 3 images of the steel plate surface in the corresponding area, and the images need to be uploaded for blur detection after shooting. If they are not clear, the acquisition continues until at least three high-definition and clear pictures are obtained. At the same time, the pictures taken by the high-definition camera in each area are cropped according to their regional positions and then compared and analyzed to ensure that the final analysis results of all images can accurately correspond to the corresponding areas.
[0012] In the steel plate flaw detection method as described above, in step three, at least two ultrasonic straight probes and two ultrasonic oblique probes are set in each area.
[0013] In the steel plate flaw detection method as described above, each ultrasonic straight probe and ultrasonic oblique probe in the step three is respectively subjected to three ultrasonic flaw detection tests and all the obtained ultrasonic spectra are subsequently tested. If there is an area with less than one-half of abnormal ultrasonic spectra, the area is re-ultrasonic flaw detection. If there is an area with more than one-half of abnormal ultrasonic spectra, the area is re-ultrasonic flaw detection and marked as a suspected area of key steel plate defects.
[0014] According to the above-mentioned steel plate flaw detection method, the construction operation of the steel plate defect image set in the step 4 is as follows: the defects that have appeared on the surface of the steel plate are counted and classified, and then 100 images with obvious defect features are selected from each category and collected into the steel plate defect image set, and then the feature image points of the 100 images with obvious defect features selected from each category are extracted and recorded, and then 500 defect images of different sizes and depths are selected from each category to train the steel plate defect image set through a convolutional neural network and judge the analysis accuracy of the steel plate defect image set according to the training results. If the accuracy is not enough, the training defect image with errors is added to the corresponding category of the steel plate defect image set, and then the feature image points are extracted and trained again until the accuracy is reached. The standard is met, and the image is collected into the steel plate defect image set. The image feature points of the input image of each area are extracted and compared. The name of the corresponding classification defect is output when the defect meets the classification, so as to quickly determine the type of steel plate defect, and all images in each area are analyzed and compared. Only when the feature image points of more than two-thirds of the image are matched with the feature image points in the steel plate defect image set, it will be determined that there are surface defects in the area. If there are no more than two-thirds of the feature image points of the image that are matched with the feature image points in the steel plate defect image set, the corresponding area labeling is questionable, and it can be judged by repeatedly comparing and detecting the image of the area and re-collecting the image. At the same time, the image can be manually enlarged to determine whether there are surface defects.
[0015] In the above-mentioned steel plate flaw detection method, the construction operation of the steel plate defect ultrasonic spectrum set in the step 4 is: statistics and classification of defects previously occurring inside the steel plate, and then 200 steel plate defect ultrasonic spectrums are selected for each classification and collected into the steel plate defect ultrasonic spectrum set, and then 100 abnormal ultrasonic spectrums are selected for each classification, and the abnormal ultrasonic spectrum is the spectrum of the steel plate that is finally detected to be defect-free but the spectrum is abnormal, and then characteristic peaks are respectively extracted and stored in the steel plate defect ultrasonic spectrum set, and then 500 steel plate defect ultrasonic spectrums that are finally confirmed to have defects and 200 abnormal ultrasonic spectrums are selected for each classification, and the steel plate defect ultrasonic spectrum set is trained by a convolutional neural network, and the analysis accuracy of the steel plate defect image set is judged according to the training results, so that the spectrum with abnormalities can be re-analyzed. The analysis is repeated to determine whether it is a real steel plate defect or just an abnormality in the spectrum and the steel plate has no defects. The ultrasonic spectrum with defects can be conveniently classified. At the same time, if the accuracy is not enough, the training ultrasonic spectrum with errors is added to the corresponding classification of the steel plate defect ultrasonic spectrum set, and then the characteristic image points are extracted and trained again until the accuracy is met; and all ultrasonic spectra in each area are analyzed and compared. Only when more than two-thirds of the characteristic peaks of the ultrasonic spectrum are consistent with the characteristic peaks in the steel plate defect ultrasonic spectrum set, will it be determined that there are surface defects in the area. If no more than two-thirds of the characteristic peaks of the ultrasonic spectrum are consistent with the characteristic peaks in the steel plate defect ultrasonic spectrum set, the corresponding area marking is questionable, which can be judged by repeatedly comparing and detecting the ultrasonic spectrum of the area and re-collecting the ultrasonic spectrum.
[0016] In a steel plate flaw detection method as described above, when the thickness of the steel plate is relatively thin, the entire roll of rolled steel needs to be continuously inspected, and a labeling device is set at the entrance of the first steel plate flaw detection device to perform continuous labeling each time it moves forward, thereby distinguishing the steel plates for each regional classification inspection by labeling, so as to facilitate the subsequent search when processing the defective parts of the steel plate.
[0017] In the steel plate flaw detection method as described above, the marking device uses quick-drying spray paint material for marking, and fluorescent powder can be added to ensure that the marking boundary position can be quickly found.
[0018] The advantages of the present invention are: the present invention collects images through the first steel plate flaw detection device, obtains ultrasonic maps through the second steel plate flaw detection device, and determines whether there are surface defects and internal defects after comparing the images with the steel plate defect image set and the ultrasonic maps with the steel plate defect ultrasonic map set. The detection result has a high accuracy rate, which avoids the situation where there are no defects after processing, resulting in reduced production efficiency. DETAILED DESCRIPTION
[0019] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described in combination with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0020] A steel plate flaw detection method comprises the following steps:
[0021] Step 1: Clean the surface of the steel plate to be tested to ensure that there is no foreign matter attached to the surface of the steel plate;
[0022] Step 2: The steel plate to be inspected after being cleaned in step 1 is transported to the first steel plate flaw detection device through a transport roller. The first steel plate flaw detection device is divided into several areas. Each area of the first steel plate flaw detection device is provided with a high-definition camera, and the high-definition camera collects images of the steel plate to be inspected;
[0023] Step 3: After the first steel plate flaw detection device in step 2 completes image acquisition, it is sent to the second steel plate flaw detection device through a transport roller. The second steel plate flaw detection device is also divided into several areas, and the number and area of the areas are the same as those of the first steel plate flaw detection device. An ultrasonic probe is set in each area of the second steel plate flaw detection device to perform ultrasonic flaw detection on the steel plate to be detected;
[0024] Step 4: Compare and analyze the images collected by the first steel plate flaw detection device in each area with the images in the steel plate defect image set, and send them to the second steel plate flaw detection device for detection after the detection results are obtained. The ultrasonic spectrum obtained by the ultrasonic probe in each area of the second steel plate flaw detection device is analyzed, and the defect characteristic peaks of the abnormal ultrasonic spectrum and the steel plate defect ultrasonic spectrum in the steel plate defect ultrasonic spectrum set are analyzed and compared to determine the type of steel plate defect, and then the image analysis results of each area and the ultrasonic spectrum analysis results of each area are combined to complete the judgment of steel plate flaw detection.
[0025] Specifically, the areas divided in the first steel plate flaw detection device and the second steel plate flaw detection device in step 2 and step 3 described in this embodiment correspond to the same steel plate positions to be detected, the number of areas divided in the first steel plate flaw detection device and the second steel plate flaw detection device is 9-36, and the length range of the areas divided in the first steel plate flaw detection device and the second steel plate flaw detection device is 60-100 cm, and the width range is 60-100 cm.
[0026] Specifically, in step 2 described in this embodiment, the high-definition camera in each area takes at least three images of the steel plate surface in the corresponding area, and the images need to be uploaded for blur detection after shooting. If they are not clear, the acquisition continues until at least three high-definition and clear pictures are obtained. At the same time, the images taken by the high-definition camera in each area will be cropped according to their regional positions and then compared and analyzed to ensure that the final analysis results of all images can accurately correspond to the corresponding areas.
[0027] More specifically, in step three of this embodiment, at least two ultrasonic straight probes and two ultrasonic oblique probes are set in each area.
[0028] To be more specific, each ultrasonic straight probe and ultrasonic oblique probe in step three described in this embodiment performs three ultrasonic flaw detection tests respectively and performs subsequent tests on all the obtained ultrasonic spectra. If there is an area with less than one-half of abnormal ultrasonic spectra, the ultrasonic flaw detection is re-performed on the area. If there is an area with more than one-half of abnormal ultrasonic spectra, the ultrasonic flaw detection is re-performed on the area and the area is marked as a suspected area of key steel plate defects.
[0029] Furthermore, the construction operation of the steel plate defect image set in step four of this embodiment is as follows: statistics are collected on the defects that have appeared on the surface of the steel plate before and classified, and then 100 images with obvious defect features are selected from each category and collected into the steel plate defect image set, and then feature image points are extracted and recorded for the 100 images with obvious defect features selected from each category, and then 500 defect images of different sizes and depths are selected from each category to train the steel plate defect image set through a convolutional neural network and the analysis accuracy of the steel plate defect image set is judged according to the training results. If the accuracy is not enough, the training defect image with errors is added to the corresponding category of the steel plate defect image set, and then the feature image points are extracted and trained again until the accuracy meets the standard. At the same time, the image is collected into the steel plate defect image set, and then the image feature points are extracted from the input image of each area. When compared, the name of the corresponding classification defect is output if it meets the defect classification, so as to quickly determine the type of steel plate defect, and all images in each area are analyzed and compared. Only when the feature image points of more than two-thirds of the image are matched with the feature image points in the steel plate defect image set, it will be determined that there are surface defects in the area. If there are no more than two-thirds of the feature image points of the image that are matched with the feature image points in the steel plate defect image set, the corresponding area labeling is questionable, and it can be judged by repeatedly comparing and detecting the image of the area and re-collecting the image. At the same time, the image can be manually enlarged to determine whether there are surface defects.
[0030] Furthermore, the construction operation of the steel plate defect ultrasonic atlas set in step four of this embodiment is as follows: statistics are collected on the defects that have occurred inside the steel plate before and classified, and then 200 steel plate defect ultrasonic atlases are selected from each classification and collected into the steel plate defect ultrasonic atlas set, and then 100 abnormal ultrasonic atlases are selected from each classification, and the abnormal ultrasonic atlases are the atlases that are finally detected to have no defects in the steel plate but have abnormal atlases, and then characteristic peaks are extracted and stored in the steel plate defect ultrasonic atlas set, and then 500 steel plate defect ultrasonic atlases that are finally confirmed to have defects and 200 abnormal ultrasonic atlases are selected from each classification to train the steel plate defect ultrasonic atlas set through a convolutional neural network, and the analysis accuracy of the steel plate defect image set is judged according to the training results, so that the abnormal atlases can be re-analyzed. It is analyzed and judged whether it is a real steel plate defect or just an abnormal spectrum and the steel plate has no defects. The ultrasonic spectrum with defects can be conveniently classified. At the same time, if the accuracy is not enough, the training ultrasonic spectrum with errors is added to the corresponding classification of the steel plate defect ultrasonic spectrum set, and then the characteristic image points are extracted and trained again until the accuracy is met; and all ultrasonic spectra in each area are analyzed and compared. Only when more than two-thirds of the characteristic peaks of the ultrasonic spectrum are matched with the characteristic peaks in the steel plate defect ultrasonic spectrum set, it will be determined that there are surface defects in the area. If no more than two-thirds of the characteristic peaks of the ultrasonic spectrum are matched with the characteristic peaks in the steel plate defect ultrasonic spectrum set, the corresponding area marking is questionable, which can be judged by repeatedly comparing and detecting the ultrasonic spectrum of the area and re-collecting the ultrasonic spectrum.
[0031] Furthermore, when the thickness of the steel plate described in this embodiment is relatively thin, the entire roll of rolled steel needs to be continuously inspected, and a labeling device is provided at the entrance of the first steel plate flaw detection device to perform continuous labeling each time it advances, thereby distinguishing the steel plates for each regional classification inspection by labeling, thereby facilitating the subsequent search for defective parts of the steel plate when processing them.
[0032] Furthermore, the marking device described in this embodiment uses quick-drying spray paint material for marking, and fluorescent powder can be added to ensure that the marking boundary position can be quickly found.
[0033] Through the final training, the present invention can detect the surface defects of steel plates with an accuracy of 99.98% by image, and the internal defects of steel plates with an accuracy of 99.92% by ultrasonic spectrum, thereby avoiding the situation of no defects after processing and effectively improving production efficiency.
[0034] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A steel plate flaw detection method, characterized in that: The steps include: Step 1: Clean the surface of the steel plate to be tested to ensure that there is no foreign matter attached to the surface of the steel plate; Step 2: The steel plate to be inspected after being cleaned in step 1 is transported to the first steel plate flaw detection device through a transport roller. The first steel plate flaw detection device is divided into several areas. Each area of the first steel plate flaw detection device is provided with a high-definition camera, and the high-definition camera collects images of the steel plate to be inspected; Step 3: After the first steel plate flaw detection device in step 2 completes image acquisition, it is sent to the second steel plate flaw detection device through a transport roller. The second steel plate flaw detection device is also divided into several areas, and the number and area of the areas are the same as those of the first steel plate flaw detection device. An ultrasonic probe is set in each area of the second steel plate flaw detection device to perform ultrasonic flaw detection on the steel plate to be detected; Step 4: Compare and analyze the images collected by the first steel plate flaw detection device in each area with the images in the steel plate defect image set, and send them to the second steel plate flaw detection device for detection after the detection results are obtained. The ultrasonic spectrum obtained by the ultrasonic probe in each area of the second steel plate flaw detection device is analyzed, and the defect characteristic peaks of the abnormal ultrasonic spectrum and the steel plate defect ultrasonic spectrum in the steel plate defect ultrasonic spectrum set are analyzed and compared to determine the type of steel plate defect, and then the image analysis results of each area and the ultrasonic spectrum analysis results of each area are combined to complete the judgment of steel plate flaw detection.
2. A steel plate flaw detection method according to claim 1, characterized in that: The areas divided in the first steel plate flaw detection device and the second steel plate flaw detection device in step 2 and step 3 correspond to the same steel plate positions to be detected, the number of areas divided in the first steel plate flaw detection device and the second steel plate flaw detection device is 9-36, and the length range of the areas divided in the first steel plate flaw detection device and the second steel plate flaw detection device is 60-100 cm, and the width range is 60-100 cm.
3. A steel plate flaw detection method according to claim 1, characterized in that: In the step 2, the high-definition camera in each area takes at least 3 images of the steel plate surface in the corresponding area, and the images need to be uploaded for blur detection after shooting. If they are not clear, the acquisition continues until at least three high-definition and clear pictures are obtained. At the same time, the images taken by the high-definition camera in each area are cropped according to their regional positions and then compared and analyzed to ensure that the final analysis results of all images can accurately correspond to the corresponding areas.
4. A steel plate flaw detection method according to claim 1, characterized in that: In the step three, at least two ultrasonic straight probes and two ultrasonic oblique probes are set in each area.
5. A steel plate flaw detection method according to claim 4, characterized in that: Each ultrasonic straight probe and ultrasonic oblique probe in the step three is subjected to three ultrasonic flaw detection tests respectively, and all the obtained ultrasonic spectra are subsequently tested. If there is an area with less than one-half of abnormal ultrasonic spectra, the area is re-ultrasonic flaw detection. If there is an area with more than one-half of abnormal ultrasonic spectra, the area is re-ultrasonic flaw detection and marked as a suspected area of key steel plate defects.
6. A steel plate flaw detection method according to claim 1, characterized in that: The construction operation of the steel plate defect image set in the step 4 is as follows: the defects that have appeared on the surface of the steel plate are counted and classified, and then 100 images with obvious defect features are selected from each category and collected into the steel plate defect image set, and then the feature image points of the 100 images with obvious defect features selected from each category are extracted and recorded, and then 500 defect images of different sizes and depths are selected from each category to train the steel plate defect image set through a convolutional neural network and judge the analysis accuracy of the steel plate defect image set according to the training results. If the accuracy is not enough, the training defect image with errors is added to the corresponding category of the steel plate defect image set, and then the feature image points are extracted and trained again until the accuracy meets the standard. At the same time, the image The steel plate defect image set is collected and the image feature points of the input images of each area are extracted. When compared, the name of the corresponding classification defect is output for the defect that meets the classification, so as to quickly determine the type of steel plate defect, and all images in each area are analyzed and compared. Only when the feature image points of more than two-thirds of the images are matched with the feature image points in the steel plate defect image set, it will be determined that there are surface defects in the area. If there are no more than two-thirds of the feature image points of the images that are matched with the feature image points in the steel plate defect image set, the corresponding area labeling is questionable, and it can be judged by repeatedly comparing and detecting the images of the area and re-collecting the images. At the same time, it can also be determined by manually zooming in on the image to determine whether there are surface defects.
7. A steel plate flaw detection method according to claim 1, characterized in that: The construction operation of the steel plate defect ultrasonic atlas set in the step 4 is as follows: the defects that have appeared inside the steel plate before are counted and classified, and then 200 steel plate defect ultrasonic atlases are selected for each classification and included in the steel plate defect ultrasonic atlas set, and then 100 abnormal ultrasonic atlases are selected for each classification, and the abnormal ultrasonic atlas is the atlas of the steel plate that is finally detected to have no defects but the atlas is abnormal, and then the characteristic peaks are respectively extracted and stored in the steel plate defect ultrasonic atlas set, and then 500 steel plate defect ultrasonic atlases that are finally confirmed to have defects and 200 abnormal ultrasonic atlases are selected for each classification to train the steel plate defect ultrasonic atlas set through a convolutional neural network and the analysis accuracy of the steel plate defect image set is judged according to the training results, so that the abnormal atlas can be analyzed again to judge its Is it a real steel plate defect or just an abnormal spectrum and the steel plate has no defects? The ultrasonic spectrum with defects can be conveniently classified. At the same time, if the accuracy is not enough, the training ultrasonic spectrum with errors is added to the corresponding classification of the steel plate defect ultrasonic spectrum set, and then the characteristic image points are extracted and trained again until the accuracy is met; and all ultrasonic spectra in each area are analyzed and compared. Only when more than two-thirds of the characteristic peaks of the ultrasonic spectrum are consistent with the characteristic peaks in the steel plate defect ultrasonic spectrum set, it will be determined that there are surface defects in the area. If no more than two-thirds of the characteristic peaks of the ultrasonic spectrum are consistent with the characteristic peaks in the steel plate defect ultrasonic spectrum set, the corresponding area marking is questionable, which can be judged by repeatedly comparing and detecting the ultrasonic spectrum of the area and re-collecting the ultrasonic spectrum.
8. A steel plate flaw detection method according to claim 1, characterized in that: When the thickness of the steel plate is relatively thin, the entire rolled steel coil needs to be continuously inspected. A labeling device is set at the entrance of the first steel plate flaw detection device to perform continuous labeling each time it moves forward, thereby distinguishing the steel plates for each regional classification inspection, thereby facilitating the search when the defective parts of the steel plate are subsequently processed.
9. A steel plate flaw detection method according to claim 8, characterized in that: The marking device uses quick-drying spray paint material for marking, and fluorescent powder can be added to ensure that the marking boundary position can be quickly found.
Citation Information
Cited By
Large steel structure damage detection method and system based on intelligent sensor
CN121090534A
