Method for detecting ridge line breakage in crystal pulling process, detection device applying method and electronic equipment

Abstract

The invention relates to a method for detecting ridge line breakage in a crystal pulling process, a detection device using the method and electronic equipment, and adopts the technical scheme that an image to be detected in a monocrystalline silicon equal-diameter process is obtained, and the image to be detected is processed by using a line breakage detection model; and detecting whether a broken line exists in the area where the monocrystalline silicon ridge line is located in the image to be detected through the sample label representing whether the broken line exists in the area where the monocrystalline silicon ridge line is located in the image to be detected of the broken line detection model, and forming broken line alarm information for alarming. The broken line detection method has the advantages that the broken line detection model with the machine learning model capability is used for learning the broken line characteristics of the ridge line, whether the broken line occurs on the single crystal silicon ridge line or not can be rapidly detected through the broken line detection model, and therefore the broken line detection efficiency of the single crystal silicon ridge line is improved.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor material processing technology, and in particular to a method for detecting edge line breakage in a crystal pulling process, and a detection device and electronic equipment using the method. Background Art

[0002] Crystal pulling refers to melting high-purity silicon raw materials at high temperature and immersing single-crystal silicon seed crystals into the melt. With the control of temperature and the pulling of the seed crystals, the melt gradually solidifies around the seed crystals and grows into a complete single-crystal silicon rod. This method can produce high-purity, high-quality single-crystal silicon materials, which are widely used in semiconductors and solar energy.

[0003] In the later stage of the crystal pulling process, the crystal rod will grow with equal diameter, slowly rotate at the liquid surface and grow into a cylindrical solid crystal rod, and vertical ridges will be formed on its side surface. During the production process, due to some factors (such as substandard environment, mismatch of crystal pulling speed or improper operation of personnel, etc.), the ridges will break or disappear. This situation is called broken ridges (or broken lines). The crystal rods produced after the broken ridges are unqualified. Therefore, if the broken ridges are not discovered in time, the crystal rods grown at the end will become waste, resulting in a waste of cost.

[0004] When detecting broken edges during the equal diameter growth process of traditional single crystal furnaces, manual inspection is often used to judge whether the single crystal silicon edges are broken. This method is prone to have a certain impact on the output and yield rate due to differences in people's skill levels, low detection efficiency, and different judgment standards. Summary of the invention

[0005] The purpose of the present invention is to solve the above-mentioned problems existing in the prior art and to provide a method for detecting broken edge lines in a crystal pulling process, a detection device and an electronic device using the same, to obtain an image to be tested in a process of equal diameter of single crystal silicon, and to process the image to be tested using a broken line detection model with machine learning model capabilities. The area where the single crystal silicon edge line is located in the image to be tested of the broken line detection model has a sample label representing whether there is a broken line, so as to detect whether there is a broken line in the area where the single crystal silicon edge line is located in the image to be tested, and to generate broken line alarm information for alarm, thereby helping to improve the efficiency of single crystal silicon edge break detection.

[0006] The above technical objectives of the present invention are mainly solved by the following technical solutions:

[0007] The technical solution of the first technical subject of the present invention is as follows:

[0008] A method for detecting edge line breakage during crystal pulling, characterized in that the method comprises the following steps:

[0009] Obtaining several images to be measured during the process of equal diameter of single crystal silicon;

[0010] Based on the image to be tested, a wire break detection model is obtained by training a deep neural network machine learning model;

[0011] Processing the image to be tested based on the broken wire detection model to obtain a detection result;

[0012] The detection result indicates whether there is a broken line in the area where the single crystal silicon edge line is located in the image to be tested;

[0013] Based on the detection result, line break alarm information is obtained, and the line break alarm information is used to indicate that the single crystal silicon equal-diameter crystal pulling process is to be processed.

[0014] As a further improvement and supplement to the above technical scheme, the present invention adopts the following technical measures: the image to be tested has one or several sample areas; each of the sample areas is the area where the edge line on the single crystal silicon is located, and each sample area corresponds to a sample label, and the sample label characterizes whether there is a broken edge line in the corresponding sample area. Based on the broken line detection model, the detection result characterizes whether there is a broken edge line in the area where the single crystal silicon edge line in the image to be tested is located.

[0015] Preferably, according to the detection result, the step of obtaining disconnection alarm information comprises:

[0016] Collect a number of images to be tested of a set number of recent times;

[0017] Obtaining a corresponding number of the test results;

[0018] Determine the number of broken lines detected in each of the detection results and the total number of detections to obtain a number of edge line broken rates;

[0019] The disconnection alarm information is generated based on the edge disconnection rate.

[0020] Preferably, the detection result, the step of obtaining disconnection alarm information comprises:

[0021] Collecting a number of images to be tested in a time sequence within a recent period of time;

[0022] Obtaining a corresponding number of the detection results;

[0023] Determine the number of broken lines detected in each of the detection results and the total number of detections to obtain a number of edge line broken rates;

[0024] The disconnection alarm information is generated based on the edge disconnection rate.

[0025] Preferably, the detection frequency is set, and when a new image is acquired from the image to be detected for detection, the image at the new moment is used as the current image, and the steps include:

[0026] Detect the new image in the image to be tested, obtain the detection result at the new moment, take the image at the new moment as the new current image, and then calculate the current edge break rate corresponding to the new image according to the same rule, and so on, until the corresponding current edge break rates are calculated for all the images to be tested.

[0027] Preferably, when the edge line break rate is less than or equal to the break rate threshold, the break count value is set to 0;

[0028] When the edge line breakage rate is greater than the breakage rate threshold, the breakage count value increases by 1;

[0029] If the edge line break rate is less than or equal to the break rate threshold, the break count value is reset and recalculated;

[0030] The sizes of the edge line breakage rates and the breakage rate thresholds are judged in sequence to obtain a final breakage count value, and when the final breakage count value is greater than or equal to the count threshold, a breakage alarm message is generated.

[0031] Preferably, the images to be tested have an image sequence, and the image sequence is consistent with the sequence in which the images to be tested are collected.

[0032] Preferably, the step of obtaining the image to be tested includes:

[0033] The initial image is the image collected during the single crystal furnace equal diameter process;

[0034] Obtaining annotation data for the initial image;

[0035] The labeled data includes a labeled sample area and a sample label corresponding to the sample area;

[0036] The sample label is added to the sample area in the initial image to obtain an image to be tested.

[0037] The technical solution of the second technical subject of the present invention is: a detection device based on a method for detecting edge line breakage during crystal pulling, characterized by comprising:

[0038] An image acquisition unit, used to obtain the image to be measured during the process of isodiameter of single crystal silicon;

[0039] An image detection unit processes the image to be tested based on the image to be tested and a wire break detection model obtained by training a deep neural network machine learning model to obtain a detection result corresponding to the image to be tested, wherein the detection result indicates whether there is a wire break in the area where the single crystal silicon edge line is located in the image to be tested;

[0040] A wire break alarm unit: based on the detection result, used to instruct the processing of the single crystal silicon equal-diameter crystal pulling process.

[0041] The technical solution of the third technical subject of the present invention is: an electronic device, comprising:

[0042] A storage device for storing a computer program and data generated by the operation of the computer program;

[0043] A processor is used to execute the computer program, and when the computer program is executed by the processor, the method for detecting edge line breakage during crystal pulling is implemented to achieve:

[0044] Obtaining several images to be measured during the process of equal diameter of single crystal silicon;

[0045] Based on the image to be tested, a wire break detection model is obtained by training a deep neural network machine learning model;

[0046] Processing the image to be tested based on the broken wire detection model to obtain a detection result;

[0047] The detection result indicates whether there is a broken line in the area where the single crystal silicon edge line is located in the image to be tested;

[0048] Based on the detection result, line break alarm information is obtained, and the line break alarm information is used to indicate that the single crystal silicon equal-diameter crystal pulling process is to be processed.

[0049] The present invention has the beneficial effects:

[0050] 1. After obtaining the image to be tested in the process of isodiameter of single crystal silicon, the broken line detection model is used to process the image to be tested. Since the area where the single crystal silicon edge line is located in the image to be tested of the broken line detection model has a sample label representing whether there is a broken line, the broken line detection model can detect whether there is a broken line in the area where the single crystal silicon edge line is located in the image to be tested.

[0051] 2. The real-time image of the equal diameter process is captured by a camera installed in the crystal pulling furnace to enhance the real-time performance. The characteristics of the edge line break are learned by the line break detection model with machine learning capabilities, so that whether the single crystal silicon edge line is broken can be quickly detected. Combined with multiple consecutive detections and calculations, the line break rate of the equal diameter stage is obtained, which effectively reduces the fluctuation error that may be caused by a single detection of line break, thereby improving the line break detection efficiency of the single crystal silicon edge line and improving the accuracy of line break detection.

[0052] 3. It can automatically obtain the broken wire situation in the process of equal diameter, and process and analyze it, so as to realize the automatic detection and judgment of the broken wire of the single crystal silicon edge line, which is helpful to timely discover and solve the problems that may arise in the crystal pulling process, and can not only reduce the waste of raw materials and the scrap rate of crystal rods, but also ensure the stability of the production process and the stability of product quality, which is conducive to improving the product qualification rate and production benefits, thereby helping to improve production efficiency.

[0053] 4. Simplify the operation process, realize automatic detection and control, and reduce the dependence on operators in the production process

[0054] 5. The device has a simple structure and is easy to operate, which reduces the frequency of operators entering the production workshop and operating near the crystal pulling furnace, reduces the direct contact of operators with dangerous environments such as high temperature and high pressure, and reduces the difficulty of operation, labor intensity and safety risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A schematic flow chart of a method for detecting edge line breakage in a crystal pulling process according to the present invention;

[0056] Figure 2 A schematic flow chart of an application of the method for detecting edge line breakage in a crystal pulling process according to the present invention;

[0057] Figure 3 It is a schematic diagram of an initial image of the ridgeline of the single crystal silicon equal diameter stage involved in the present invention;

[0058] Figure 4 It is a schematic diagram of the single crystal silicon equal diameter stage ridge line under normal conditions (the lower end of the ridge line almost touches the white aperture) involved in the present invention;

[0059] Figure 5 It is a schematic diagram of the case where the edge line of the single crystal silicon equal diameter stage involved in the present invention is broken (the lower end of the edge line is far away from the white aperture).

[0060] Figure 6 and Figure 7 Rectangular frames are Figure 4 and Figure 5 The above diagram shows the sample area (sample labels are represented by 0 or 1). DETAILED DESCRIPTION

[0061] The technical solution of the present invention is further specifically described below through embodiments and in conjunction with the accompanying drawings.

[0062] Example 1: Figure 1-Figure 5 As shown, the technical solution of the first technical subject of the present invention is:

[0063] A method for detecting edge line breakage during crystal pulling, characterized in that the method comprises the following steps:

[0064] Obtaining several images to be measured during the process of equal diameter of single crystal silicon;

[0065] Based on the image to be tested, a wire break detection model is obtained by training a deep neural network machine learning model;

[0066] Processing the image to be tested based on the broken wire detection model to obtain a detection result;

[0067] The detection result indicates whether there is a broken line in the area where the single crystal silicon edge line is located in the image to be tested;

[0068] Based on the detection result, line break alarm information is obtained, and the line break alarm information is used to indicate that the single crystal silicon equal-diameter crystal pulling process is to be processed.

[0069] Further improvement and supplement to the above technical solutions:

[0070] The image to be tested has one or several sample areas; each of the sample areas is the area where the edge line on the single crystal silicon is located, and each sample area corresponds to a sample label, and the sample label represents whether there is a broken edge line in the corresponding sample area. Based on the broken line detection model, the detection result represents whether there is a broken edge line in the area where the single crystal silicon edge line in the image to be tested is located.

[0071] In practical applications, according to the detection results, the steps of obtaining disconnection alarm information include:

[0072] Collect a number of images to be tested of a set number of recent times;

[0073] Obtaining a corresponding number of the test results;

[0074] Determine the number of broken lines detected in each of the detection results and the total number of detections to obtain a number of edge line broken rates;

[0075] The disconnection alarm information is generated based on the edge disconnection rate.

[0076] In practical applications, the detection result and the step of obtaining disconnection alarm information include:

[0077] Collecting a number of images to be tested in a time sequence within a recent period of time;

[0078] Obtaining a corresponding number of the detection results;

[0079] Determine the number of broken lines detected in each of the detection results and the total number of detections to obtain a number of edge line broken rates;

[0080] The disconnection alarm information is generated based on the edge disconnection rate.

[0081] In practical applications, the detection frequency is set, and when a new image is acquired from the image to be detected for detection, the image at the new moment is used as the current image, and the steps include:

[0082] Detect the new image in the image to be tested, obtain the detection result at the new moment, take the image at the new moment as the new current image, and then calculate the current edge break rate corresponding to the new image according to the same rule, and so on, until the corresponding current edge break rates are calculated for all the images to be tested.

[0083] In practical applications, when the edge line break rate is less than or equal to the break rate threshold, the break count value is set to 0;

[0084] When the edge line breakage rate is greater than the breakage rate threshold, the breakage count value increases by 1;

[0085] If the edge line break rate is less than or equal to the break rate threshold, the break count value is reset and recalculated;

[0086] The sizes of the edge line breakage rates and the breakage rate thresholds are judged in sequence to obtain a final breakage count value, and when the final breakage count value is greater than or equal to the count threshold, a breakage alarm message is generated.

[0087] In practical applications, the images to be tested have an image sequence, and the image sequence is consistent with the sequence in which the images to be tested are collected.

[0088] In practical applications, the steps of obtaining the image to be tested include:

[0089] The initial image is the image collected during the single crystal furnace equal diameter process;

[0090] Obtaining annotation data for the initial image;

[0091] The labeled data includes a labeled sample area and a sample label corresponding to the sample area;

[0092] The sample label is added to the sample area in the initial image to obtain an image to be tested.

[0093] According to practical applications, the above technical solution is further improved by changing the description method: a method for detecting edge line breakage during crystal pulling includes the following steps:

[0094] Step 1: Obtain the image to be tested during the process of isodiameter of single crystal silicon:

[0095] The image to be tested in step 1 can be captured by an industrial camera starting from the equal diameter shoulder of the single crystal silicon to obtain multiple frames of images to be tested. The image to be tested should evenly cover all processes from the beginning to the end of the equal diameter as much as possible.

[0096] The specific method is to start from the single crystal silicon with equal diameter, and to collect images of the single crystal silicon according to a preset target time length through an industrial camera, and the target time length can be s seconds. Thus, a frame of the image to be tested can be obtained every target time length. Afterwards, in this embodiment, step 2 can be performed for each frame of the image to be tested.

[0097] Step 2: Use the broken line detection model to process the image to be tested to determine whether there is a broken line in the image to be tested:

[0098] The broken wire detection model is obtained by training the image to be tested. The image to be tested has a sample area, which is the area where the single crystal silicon edge line is located, and the sample area has a sample label, which represents whether there is a broken wire in the sample area.

[0099] The broken wire detection model is a large model based on machine learning. For example, the broken wire detection model is a machine learning model based on a deep neural network. The image to be tested contains one or more sample areas, each of which is the area where the ridgeline on the single crystal silicon is located. Each sample area corresponds to a sample label, and the sample label can be represented by 0 or 1. 1 indicates that there is a broken wire in the sample area, and 0 indicates that there is no broken wire in the sample area. Based on this, the broken wire detection model can be used to detect the area where the single crystal silicon ridgeline is located in the image to be tested, and can also detect whether there is a broken wire in the target area;

[0100] It can be seen from the above technical scheme that in a method for detecting broken edges in the crystal pulling process provided in Example 1 of the present application, after obtaining the image to be tested in the process of isodiameter of single crystal silicon, the broken line detection model is used to process the image to be tested. Since the area where the single crystal silicon edge line is located in the image to be tested of the broken line detection model has a sample label representing whether there is a broken line, the broken line detection model can detect whether there is a broken line in the area where the single crystal silicon edge line is located in the image to be tested. It can be seen that in this embodiment, the characteristics of broken edges are learned through a machine learning model, and then the machine learning model can quickly detect whether the single crystal silicon edge line is broken, thereby improving the efficiency of single crystal silicon edge line break detection;

[0101] The image to be tested can be obtained in the following ways:

[0102] Obtain multiple frames of images collected from single crystal silicon during the historical equal diameter process;

[0103] For example, an industrial camera is used to collect an initial image of the single crystal silicon edge line at each target time interval during the entire process from the beginning of the equal diameter to the end of the equal diameter, such as N initial images;

[0104] In order to increase the diversity of the images to be tested, the initial images can also be enhanced. For example, one or more operations such as scaling, translation and rotation are performed on each frame of the acquired image, and the number of acquired images is expanded without changing the image content of the initial image;

[0105] After obtaining the image to be tested, start adding annotation data to the image to be tested, including the corresponding sample area and sample label;

[0106] For example, Figure 6 and Figure 7 As shown, the sample area is represented by a rectangular box, and the sample label is represented by 0 or 1.

[0107] In order to improve the accuracy of the sample, after using the broken wire detection model to process the image to be tested, it is also necessary to optimize the broken wire detection model according to the detection results.

[0108] The detection result can be compared with the annotation result corresponding to the image to be tested, and then the model parameters of the wire break detection model can be adjusted according to the comparison result through the loss function of the wire break detection model, so that the wire break detection model is optimized, thereby improving the detection accuracy of the wire break detection model for the next frame of the image to be tested;

[0109] After using the disconnection detection model to detect the image to be tested and obtaining the disconnection result, it is also necessary to analyze the detection result corresponding to the image to be tested to generate disconnection alarm information;

[0110] The specific method to obtain disconnection alarm information is as follows:

[0111] In order to improve the accuracy of the alarm and avoid the detection error of a single image to be tested, it is necessary to set a calculation length of the current line break rate of m, and calculate the current edge line break rate based on the latest m consecutive line break detection results;

[0112] Taking m as 30 as an example, the image detection results of the most recent 30 frames are recorded. In the most recent 30 frames of image detection starting from the image currently being detected, if there are a frames of images with broken lines detected and b frames of images without broken lines detected, then the current broken line rate is a / m, and so on. The current broken line rate represents the ratio of the number of frames a and m of images with broken lines detected in the most recent m frames of detection images.

[0113] Each time a new frame of image is detected, the new image is used as a benchmark, and the corresponding current disconnection rate is calculated according to the image detection results of the most recent m frames, and then compared with the disconnection rate threshold. If it is greater than the disconnection rate threshold, the disconnection count value is increased by 1. If the current disconnection rate is less than or equal to the disconnection rate threshold for a consecutive times during continuous detection, the disconnection count value is initialized to 0 and the counting starts again.

[0114] The disconnection count value represents the number of times the current disconnection rate is continuously detected to exceed the disconnection threshold. The continuous certain number a here can be customized according to needs. For example, when a is 1, the disconnection count value will be initialized to 0 and the disconnection count value will be recalculated as long as the current disconnection rate is less than or equal to the disconnection rate threshold once during continuous detection. If a is 3, the disconnection count value will be recalculated only when the current disconnection rate is less than or equal to the disconnection rate threshold for three consecutive times.

[0115] Similarly, if the broken wire count value keeps accumulating and exceeds the threshold of the broken wire count value, then the current process is terminated and a broken wire alarm message is generated to indicate that a broken wire has occurred in the equal diameter process of the single crystal silicon, prompting the user to handle it in time. At this time, the equal diameter pulling process of the single crystal silicon can be stopped synchronously, and the image to be tested is no longer collected. If the broken wire count value does not exceed the counting threshold until the detection is completed, that is, from the beginning to the end of the equal diameter process, each frame of the image to be tested collected has been calculated with the current broken wire rate, and finally at the end, the broken wire count value does not exceed the counting value threshold, then the current process is also terminated, indicating that no broken wire has occurred in the equal diameter process of the single crystal silicon.

[0116] Another detection method can be used. According to the detection results of the image to be tested in the last certain length m seconds, the current disconnection rate is calculated. For example, 1 frame per second, there are a total of m frames of detection results in the last m seconds. Among these detection results, if a frame is disconnected, the current disconnection rate is a / m.

[0117] For example, when m is 30, the detection results of the image to be tested in the last 30 seconds are obtained. Taking 1 frame per second as an example, a total of 30 frames of the image to be tested are detected in 30 seconds. The current disconnection rate is calculated in the same way. After that, every 1 second, the new time point is used as the reference, and the image detection results of the last 30 seconds are taken to calculate the current disconnection rate.

[0118] Set the disconnection count value to 0. The disconnection count value represents the number of seconds that the current disconnection rate exceeds the disconnection rate threshold. If the current disconnection rate is greater than the disconnection rate threshold, the disconnection count value is increased by 1. If the current disconnection rate is less than or equal to the count value threshold for a consecutive seconds, the disconnection count value needs to be accumulated again from 0. For example, when a is 1, it means that as long as the current disconnection rate is less than or equal to the count value threshold for 1 second, the disconnection count value needs to be accumulated again.

[0119] The disconnection rate threshold can be set according to the requirements. The disconnection rate threshold represents the maximum limit of the number of disconnections in the detection of the most recent m frames of images. If the current disconnection rate is greater than the disconnection rate threshold, it represents that the number of images with disconnections in the most recent m frames of the current image exceeds the maximum limit.

[0120] According to the broken wire count value, the alarm information is obtained. During the process of increasing the broken wire count value, if the broken wire count value is greater than the count value threshold, then the broken wire alarm information is generated to indicate that the single crystal silicon equalization process is to be processed. If all the images of all the equalization processes are detected and the broken wire count value is not greater than the count value threshold, then the process is terminated and there is no need for a broken wire alarm, indicating that there is no broken wire during the equalization process.

[0121] Embodiment 2: The technical solution of the second technical subject of the present invention: a detection device based on the method of detecting the broken edge line in the crystal pulling process, characterized by comprising:

[0122] An image acquisition unit, used to obtain the image to be measured during the process of isodiameter of single crystal silicon;

[0123] An image detection unit processes the image to be tested based on the image to be tested and a wire break detection model obtained by training a deep neural network machine learning model to obtain a detection result corresponding to the image to be tested, wherein the detection result indicates whether there is a wire break in the area where the single crystal silicon edge line is located in the image to be tested;

[0124] A wire break alarm unit: based on the detection result, used to instruct the processing of the single crystal silicon equal-diameter crystal pulling process.

[0125] The detection device based on the method of detecting the broken edge line during the crystal pulling process is used to implement a method based on detecting the broken edge line during the crystal pulling process of Example 1.

[0126] Embodiment 3: The technical solution of the third technical subject involved in the present invention: an electronic device, comprising:

[0127] A storage device for storing a computer program and data generated by the operation of the computer program;

[0128] A processor is used to execute the computer program, and when the computer program is executed by the processor, the method for detecting edge line breakage during crystal pulling described in embodiment 1 is implemented to achieve:

[0129] Obtaining several images to be measured during the process of equal diameter of single crystal silicon;

[0130] Based on the image to be tested, a wire break detection model is obtained by training a deep neural network machine learning model;

[0131] Processing the image to be tested based on the broken wire detection model to obtain a detection result;

[0132] The detection result indicates whether there is a broken line in the area where the single crystal silicon edge line is located in the image to be tested;

[0133] Based on the detection result, line break alarm information is obtained, and the line break alarm information is used to indicate that the single crystal silicon equal-diameter crystal pulling process is to be processed.

[0134] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. In the above embodiments, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting edge line breakage during crystal pulling, characterized in that The steps include: Obtaining several images to be measured during the process of equal diameter of single crystal silicon; Based on the image to be tested, a wire break detection model is obtained by training a deep neural network machine learning model; Processing the image to be tested based on the broken wire detection model to obtain a detection result; The detection result indicates whether there is a broken line in the area where the single crystal silicon edge line is located in the image to be tested; Based on the detection result, line break alarm information is obtained, and the line break alarm information is used to indicate that the single crystal silicon equal-diameter crystal pulling process is to be processed.

2. The method for detecting edge line breakage during crystal pulling according to claim 1, characterized in that The image to be tested has one or several sample areas; each of the sample areas is the area where the edge line on the single crystal silicon is located, and each sample area corresponds to a sample label, and the sample label represents whether there is a broken edge line in the corresponding sample area. Based on the broken line detection model, the detection result represents whether there is a broken edge line in the area where the single crystal silicon edge line in the image to be tested is located.

3. The method for detecting edge line breakage during crystal pulling according to claim 2, characterized in that According to the detection result, the step of obtaining disconnection alarm information includes: Collect a number of images to be tested of a set number of recent times; Obtaining a corresponding number of the test results; Determine the number of broken lines detected in each of the detection results and the total number of detections to obtain a number of edge line broken rates; The disconnection alarm information is generated based on the edge disconnection rate.

4. The method for detecting edge line breakage during crystal pulling according to claim 2, characterized in that According to the detection result, the step of obtaining disconnection alarm information includes: Collecting a number of images to be tested in a time sequence within a recent period of time; Obtaining a corresponding number of the detection results; Determine the number of broken lines detected in each of the detection results and the total number of detections to obtain a number of edge line broken rates; The disconnection alarm information is generated based on the edge disconnection rate.

5. The method for detecting edge line breakage during crystal pulling according to claim 3 or 4, characterized in that: The detection frequency is set, and when a new image is acquired from the image to be detected for detection, the image at the new moment is used as the current image, and the steps include: Detect the new image in the image to be tested, obtain the detection result at the new moment, take the image at the new moment as the new current image, and then calculate the current edge break rate corresponding to the new image according to the same rule, and so on, until the corresponding current edge break rates are calculated for all the images to be tested.

6. The method for detecting edge line breakage during crystal pulling according to claim 3 or 4, characterized in that: When the edge line break rate is less than or equal to the break rate threshold, the break count value is set to 0; When the edge line breakage rate is greater than the breakage rate threshold, the breakage count value increases by 1; If the edge line break rate is less than or equal to the break rate threshold, the break count value is reset and recalculated; The sizes of the edge line breakage rates and the breakage rate thresholds are judged in sequence to obtain a final breakage count value, and when the final breakage count value is greater than or equal to the count threshold, a breakage alarm message is generated.

7. The method for detecting edge line breakage during crystal pulling according to claim 3 or 4, characterized in that: The images to be tested have an image sequence, and the image sequence is consistent with the sequence in which the images to be tested are collected.

8. The method for detecting edge line breakage during crystal pulling according to claim 1, characterized in that: The steps of obtaining the image to be tested include: The initial image is the image collected during the single crystal furnace equal diameter process; Obtaining annotation data for the initial image; The labeled data includes a labeled sample area and a sample label corresponding to the sample area; The sample label is added to the sample area in the initial image to obtain an image to be tested.

9. A detection device based on the method of detecting edge line breakage during crystal pulling, characterized in that include: An image acquisition unit, used to obtain the image to be measured during the process of isodiameter of single crystal silicon; An image detection unit processes the image to be tested based on the image to be tested and a wire break detection model obtained by training a deep neural network machine learning model to obtain a detection result corresponding to the image to be tested, wherein the detection result indicates whether there is a wire break in the area where the single crystal silicon edge line is located in the image to be tested; A wire break alarm unit: based on the detection result, used to instruct the processing of the single crystal silicon equal-diameter crystal pulling process.

10. An electronic device comprising: A storage device for storing a computer program and data generated by the operation of the computer program; A processor, configured to execute the computer program, wherein when the computer program is executed by the processor, the method for detecting edge line breakage during crystal pulling according to any one of claims 1 to 8 is implemented to achieve: Obtaining several images to be measured during the process of equal diameter of single crystal silicon; Based on the image to be tested, a wire break detection model is obtained by training a deep neural network machine learning model; Processing the image to be tested based on the broken wire detection model to obtain a detection result; The detection result indicates whether there is a broken line in the area where the single crystal silicon edge line is located in the image to be tested; Based on the detection result, line break alarm information is obtained, and the line break alarm information is used to indicate that the single crystal silicon equal-diameter crystal pulling process is to be processed.

Images