A method for calculating the arcing position and time of low-clutter antennas using machine learning

Abstract

The present invention discloses a method for using machine learning to statistically calculate the arcing position and duration of a low-clutter antenna. The method relates to the field of magnetic confinement fusion. During operation of a magnetic confinement fusion device, a visible camera is used to observe the bright spots produced by arcing, continuously recording the changing process of the arcing state of the low-clutter antenna. A large number of different arcing states are identified and classified, and a corresponding database is established. The database is used as input in a deep convolutional neural network analysis program, which statistically learns and classifies the different arcing states. The results of manual and program identification of the arcing states are compared, and the accuracy of the program calculations is verified while calibrating. Based on the program identification results, the machine learning-based DBSCAN algorithm is further used to normalize and cluster the classified images of the different arcing states, and the duration of different arcing positions on the low-clutter antenna is calculated. The present invention can achieve real-time identification of arcing states for low-clutter antennas and simultaneously calculate their arcing positions and durations.

Description

Technical Field

[0001] The present invention belongs to the field of magnetic confinement nuclear fusion, and in particular relates to a method for using machine learning to calculate the arcing position and time of a low-clutter antenna. Background Art

[0002] In a tokamak, plasma toroidal currents play a crucial role in plasma confinement. However, conventional methods use ohmic induction to drive toroidal currents. The limited variable magnetic flux provided by the transformer limits the duration of toroidal current maintenance, necessitating the use of complementary non-inductive methods to drive the plasma current. Low-hybrid current drive (LHCD) is a highly efficient non-inductive current drive method currently used in tokamaks and a key approach to achieving long-pulse steady-state operation. LHCD can be used to maintain the plasma current. However, arcing in tokamaks' low-hybrid antennas is prone to arcing during high-power injection. This phenomenon creates localized high-density plasma on the antenna surface, reducing low-hybrid coupling efficiency and impacting the stable operation of the low-hybrid system. Arcing also generates a large number of high-atomic-number metal impurities, which contaminate the plasma, severely impacting the stability of the plasma discharge, and even causing plasma rupture, compromising the safe operation of the tokamak. Therefore, understanding the mechanism of arcing on the surface of low-hybrid antennas and enabling real-time monitoring and identification are crucial for the stable operation of tokamaks. Due to the varying timing of the low-clutter antenna's heating power and the varying plasma current phases during discharge, the low-clutter antenna inevitably experiences varying degrees of arcing. This arcing state also varies with the varying discharge conditions of each shot. Currently, low-clutter antennas experience varying degrees of arcing, which can severely impact the steady-state operation of tokamaks. Manually counting these arcing states, locations, and times consumes significant human resources and will be difficult to achieve in real time during future device operation. Therefore, a convenient, fast, and accurate method for counting the location and timing of arcing in low-clutter antennas is urgently needed. Summary of the Invention

[0003] To address the aforementioned technical issues and more accurately measure the impact of different arcing states on low-clutter antennas, the present invention provides a method for using machine learning to measure the arcing position and duration of low-clutter antennas. These different arcing states are observed and captured using a high-speed CCD visible camera. The output of a deep convolutional neural network (DCNN) training program is then compared with existing manually determined arcing states that are not recorded in a database to accurately identify the different arcing states. Based on the program output, a machine learning-based DBSCAN algorithm is further employed to perform statistical analysis of the arcing position and duration. The trained machine learning program is then applied to measure the different arcing states, including arcing position and duration, in real time during the device's discharge.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions:

[0005] A method for using machine learning to calculate the arcing position and time of a low-clutter antenna includes the following steps:

[0006] Step 1: Before the magnetic confinement nuclear fusion device is operated, adjust the high-speed CCD visible camera and the endoscope system at the front end of the visible light path, set the exposure time and acquisition frame rate of the high-speed CCD visible camera according to experimental conditions, and realize the observation of the arcing phenomenon of the low-clutter antenna;

[0007] Step 2: When the magnetic confinement nuclear fusion device is in operation, a high-speed CCD visible camera is used to record the entire process of the changing image of the bright spot state generated when the low-clutter antenna arcs;

[0008] Step 3: In combination with the changes in the operating parameters of the magnetic confinement nuclear fusion device, an image database including three different arc ignition states is established using the images of the changing process; the three different arc ignition states are a non-arcing state, a single-point arc ignition state, and a multi-point arc ignition state;

[0009] Step 4: Use the image database established in step 3 as the input of the deep convolutional neural network model to train the deep convolutional neural network model. Combined with the image output by the deep convolutional neural network model, compare and analyze the manually determined arc starting states that are not recorded in the image database, accurately identify different arc starting states, and obtain output results.

[0010] Step 5: For different arcing states in the output result of step 4, a machine learning-DBSCAN algorithm is used to count all arcing positions on the low-clutter antenna and the arc duration at the arcing position during each shot discharge process;

[0011] Step 6: When the low-clutter antenna arcs, the images collected during one or two discharge processes are used as input to the machine learning-DBSCAN algorithm. The output calculation results are compared with the manual statistical results to achieve accurate statistics of the arcing position and duration by the machine learning-DBSCAN algorithm.

[0012] Step 7. Apply the deep convolutional neural network model and the machine learning-DBSCAN algorithm to the discharge process of the magnetic confinement fusion device. Combined with the experimental control conditions during the operation of the magnetic confinement fusion device, after achieving accurate identification of different arcing states, the machine learning-DBSCAN algorithm is further used to realize the statistics of arcing position and duration.

[0013] During the operation of a magnetic confinement fusion device, the present invention uses a high-speed CCD visible camera to observe bright spots generated by arcing, thereby continuously recording the changing process of the arcing state of a low-clutter antenna. A large number of different arcing states are identified and classified, and a corresponding image database is established. The image database is used as input to a deep convolutional neural network, and statistical learning is used to classify different arcing states. The results of manual identification and deep convolutional neural network identification of arcing states are compared. The accuracy of the deep convolutional neural network calculation is verified while calibrating. Based on the results of the deep convolutional neural network identification, a machine learning-DBSCAN algorithm is further used to normalize and cluster the classified images for different arcing states, and statistics are collected on different arcing positions and durations on the low-clutter antenna. The image database is ultimately applied to the real-time discharge process of the magnetic confinement fusion device.

[0014] Beneficial effects:

[0015] This method accurately and in real time determines the different arcing states within each shot discharge, as well as the arcing locations and arc durations at those locations. This provides more accurate data analysis for studying the coupling performance of low-clutter antennas and the physical operation of the device. This method is of great significance for further studying their impact on plasma operation and developing methods for controlling arc triggering, ensuring the steady-state operation of low-clutter current drive systems, and promoting the development of long-pulse steady-state operation in tokamaks. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flow chart of the method for using machine learning to calculate the real-time statistics of the arcing position and time of a low-clutter antenna according to the present invention;

[0017] Figure 2 This is a schematic diagram of the low-clutter antenna in the non-arcing state;

[0018] Figure 3 This is a schematic diagram of the single-point arc starting state of the low-clutter antenna;

[0019] Figure 4 This is a schematic diagram of the multi-point arcing state of the low-clutter antenna;

[0020] Figure 5 This is a statistical diagram of the number of images of different arcing states under a complete discharge of a gun that have not been recorded in the database and have been recognized by the output of the deep convolutional neural network program;

[0021] Figure 6 The deep convolutional neural network program and machine learning-DBSCAN are used to collect real-time statistics of the arcing position and time of the low-noise antenna during the steady-state operation of the magnetic confinement fusion device. DETAILED DESCRIPTION

[0022] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to illustrate the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0023] In order to better illustrate how to use a deep convolutional neural network to identify the arcing state of a low-clutter antenna in real time, the present invention is described in detail below with reference to the accompanying drawings and implementation cases.

[0024] like Figure 1 As shown, the arcing of the low-clutter antenna is a localized discharge phenomenon with high current density and short duration that occurs on the surface of the plasma facing the component material during discharge in a magnetic confinement fusion device. A method for using machine learning to calculate the position and time of arcing of the low-clutter antenna in an embodiment of the present invention includes the following steps:

[0025] Step 1: Before the magnetic confinement fusion device experiment, adjust the high-speed CCD visible camera and the endoscope system at the front end of the visible light path to observe the position of the low-clutter antenna;

[0026] Step 2: A high-speed CCD visible camera records the arcing phenomenon of the low-clutter antenna in the magnetic confinement fusion device and establishes a database. According to the different operating conditions of the magnetic confinement fusion device, a frame rate of 1000fps or above and an exposure time adjustment of 1-10us can be achieved. After collecting images in different states, the conditions of the low-clutter antenna in different arcing states (i.e. Figure 1 The camera in the image captures images in different arc starting states);

[0027] Step 3: Combine the low noise phase and power injection time during long pulse discharge to artificially establish different arcing conditions ( Figure 2-Figure 4 ) database; it can be roughly divided into three categories: one is the state before arcing (no power is injected - state "I"), another is the state of single-point arcing after power injection - state "II", and the last one is the state of multiple-point arcing - state "III";

[0028] Step 4: If Figure 1 As shown in the figure, a database of different arcing states is used as input, and a deep convolutional neural network learning method is used to establish a training model. The output results are compared and analyzed with images of other arcing states that are not recorded in the database, enabling the deep convolutional neural network to accurately count the different arcing states.

[0029] The input of the deep convolutional neural network learning is input into the database of manually classified different states. Its fixed size JPG (pixel 768 736) image, convolving the input image with convolution kernels to extract local features. These convolution kernels slide across the entire image to generate feature maps. The convolved feature maps undergo a nonlinear transformation using the ReLU activation function. Pooling is then performed to reduce the feature map's size. Local response normalization is performed to improve training stability. The convolution, activation, pooling, and normalization operations are repeated to gradually extract higher-level image features at different scales. Following the convolution and pooling layers, the neural network will contain several fully connected layers, which integrate the features extracted from the convolution and pooling layers and perform classification through one or more fully connected layers. The final layer of the model typically consists of a Softmax layer that selects the class with the highest probability and combines it with an argmax operation to directly obtain the final classification result. At this point, the neural network has learned the image features of all different arcing states. A dictionary is used to count the number of images in each class and output each category. When images of different arcing states of a complete gun discharge phase are introduced, the output will display the calculated different arcing states. The program outputs different arcing status results and the data in the input port database are verified and analyzed. The manually classified status in step 3 needs to be combined with other parameters, such as low-clutter antenna power injection and phase change time, and antenna port probe diagnosis, to analyze and eliminate some data with large deviations (for example, multiple bright spots generated by multiple arcing points may overlap and be judged as single-point arcing points), thereby further improving the accuracy of program recognition.

[0030] Step 5: Combine the results of the deep convolutional neural network output to identify and classify the number of various arcing states. For the classified single-point arcing state and multi-point arcing state images, the machine learning-DBSCAN algorithm without a given number of classifications is used to normalize the overall brightness of the image. The image sets a brightness threshold (preferably 150) to normalize its output. DBSCAN is a density-based clustering algorithm. Since the location of the bright spot is actually more than the size of a pixel, the distance between the two pixels needs to be given. That is, if the distance between the two points is less than this value, they will be classified into one category. After using DBSCAN to obtain the classification of the bright spot, the mean is obtained within each cluster (that is, Figure 1 The number of frames and position coordinates of the arc are detected and the brighter area, i.e., the surrounding pixels of the bright spot, are merged and marked in the image as the location where the arc is generated (i.e. Figure 1 Finally, the results are saved, including the time and number of arcs detected, the frame rate of the cine file, and the duration of the bright spot position in each shot discharge (i.e. Figure 1 ).

[0031] Step 6: When the low-clutter antenna arcs, try to use the images collected during one or two discharge processes as input to the machine learning program. Compare the output calculation results with the manual statistical results to achieve accurate statistics of the arcing location and duration by the machine learning program.

[0032] Step 7: Apply the deep convolutional neural network model and machine learning-DBSCAN algorithm to the discharge process of the magnetic confinement fusion device. Combined with the experimental control conditions during the operation of the magnetic confinement fusion device, after achieving accurate identification of different arcing states, the machine learning-DBSCAN algorithm is further used to realize the statistics of arcing position and duration.

[0033] Example:

[0034] After applying the machine program to the magnetic confinement fusion device, the low-noise antenna in the device was observed to accurately measure the arcing position and duration under different discharge conditions, such as Figure 5 、 Figure 6 As shown in the figure, the machine learning-DBSCAN algorithm is used to identify the arcing position and duration of the low-clutter antenna at all stages of operation, including:

[0035] S1: Before the magnetic confinement fusion device experiment, adjust the high-speed CCD visible camera and the endoscope system at the front end of the visible light path to achieve position observation of the low-clutter antenna;

[0036] S2: When the magnetic confinement fusion device is in stable operation, a high-speed CCD visible camera is used to capture images of the low-clutter antenna in the non-arcing state and in the multi-point and single-point arcing states at different positions;

[0037] S3: Combined with the different moments of power injection and phase change during long pulse discharge, manually classify and establish a large number of images of different arcing states and their corresponding databases; the manually classified states need to be combined with other parameters, such as low-clutter antenna power injection and phase change time, antenna port probe diagnostic information, etc. to analyze and remove some data with large deviations (such as multiple bright spots generated by multi-point arcing may overlap and be judged as single-point arcing). The image information collected during the entire stage of a certain shot of arcing is used as the input of the deep convolutional neural network program, and the output arcing states "I", "II", and "III" are shown in the following figure. Figure 5 It can be displayed on the screen (each moment corresponds to only one state). After manual verification, it is found that its statistical accuracy is high and the situation analysis of different states is accurate.

[0038] S4: Using the classified images of different arcing states and their database as the input of the deep convolutional neural network, training the deep convolutional neural network model, and comparing and analyzing the existing confirmed arcing states that are not recorded in the database at the output end, combining the results of different arcing states output by the deep convolutional neural network with the data of the input port for verification and analysis;

[0039] S5: Combined with the results of different arcing states output by the deep convolutional neural network learning program, the machine learning-DBSCAN program is further used to normalize the image and then cluster the pixels with bright spots to obtain the mean value. The number of frames and position coordinates of the arc are detected, and the brighter area, i.e., the surrounding pixels where the bright spot is located, are merged and marked in the image. This is regarded as the location where the arc is generated. The number of frames that the bright spot at this location lasts during the discharge process is recorded. The duration of the bright spot is calculated based on the parameters set by the camera. Figure 6 The image shows the location and duration of arcing after merging 10 pixels of the original image captured by a high-speed CCD visible camera of a certain gun;

[0040] S6: When the low-clutter antenna arcs, try to use the image collected during a discharge process as the input of the machine learning-DBSCAN algorithm. The output calculation results are compared and verified with the manually counted arc position and duration, so that the machine learning program can accurately count the arc position and duration.

[0041] S7: Combined with S2, the machine learning-DBSCAN algorithm is applied to the discharge process of the magnetic confinement fusion device. Taking into account the experimental control conditions during operation, the real-time identification of the arcing state and the accurate statistics of the arcing position and duration are realized, which provides data support for the precise control of the arcing state of the low-clutter antenna and promotes the steady-state operation of the low-clutter current drive system.

[0042] The present invention does not describe in detail parts that belong to the common knowledge of those skilled in the art. The above-described embodiments are merely descriptions of preferred embodiments of the present invention. The preferred embodiments do not describe all details in detail, nor do they limit the invention to the specific embodiments described. Without departing from the spirit of the present invention, various modifications and improvements made by those skilled in the art to the technical solution of the present invention should fall within the scope of protection defined by the claims of the present invention.

Claims

1. A method for using machine learning to calculate the arcing position and time of a low-clutter antenna, characterized in that: The following steps are involved: Step 1: Before the magnetic confinement nuclear fusion device is operated, adjust the high-speed CCD visible camera and the endoscope system at the front end of the visible light path, set the exposure time and acquisition frame rate of the high-speed CCD visible camera according to experimental conditions, and realize the observation of the arcing phenomenon of the low-clutter antenna; Step 2: When the magnetic confinement nuclear fusion device is in operation, a high-speed CCD visible camera is used to record the entire process of the changing image of the bright spot state generated when the low-clutter antenna arcs; Step 3: In combination with the changes in the operating parameters of the magnetic confinement nuclear fusion device, an image database including three different arc ignition states is established using the images of the changing process; the three different arc ignition states are a non-arcing state, a single-point arc ignition state, and a multi-point arc ignition state; Step 4: Use the image database established in step 3 as the input of the deep convolutional neural network model to train the deep convolutional neural network model. Combined with the image output by the deep convolutional neural network model, compare and analyze the manually determined arc starting states that are not recorded in the image database, accurately identify different arc starting states, and obtain output results. Step 5: For different arcing states in the output result of step 4, a machine learning-DBSCAN algorithm is used to count all arcing positions on the low-clutter antenna and the arc duration at the arcing position during each shot discharge process; Step 6: When the low-clutter antenna arcs, the images collected during one or two discharge processes are used as input to the machine learning-DBSCAN algorithm. The output calculation results are compared with the manual statistical results to achieve accurate statistics of the arcing position and duration by the machine learning-DBSCAN algorithm. Step 7. Apply the deep convolutional neural network model and the machine learning-DBSCAN algorithm to the discharge process of the magnetic confinement fusion device. Combined with the experimental control conditions during the operation of the magnetic confinement fusion device, after achieving accurate identification of different arcing states, the machine learning-DBSCAN algorithm is further used to realize the statistics of arcing position and duration.

2. The method for calculating the arcing position and time of a low-clutter antenna using machine learning according to claim 1, characterized in that: In step 1, the experimental conditions indicate that the duration of discharge of the magnetic confinement fusion device is different; when the magnetic confinement fusion device is in operation, the phenomenon of plasma discharge rupture is excluded from the record.

3. The method for calculating the arcing position and time of a low-clutter antenna using machine learning according to claim 1, wherein: In step 2, a high-speed CCD visible camera is used to record the different arcing states of the low-clutter antenna during each shot discharge process, and each frame of image is used for subsequent statistics of the arcing position and corresponding duration of the shot.

4. The method for calculating the arcing position and time of a low-clutter antenna using machine learning according to claim 1, wherein: In step 4, the results of deep convolutional neural network recognition and manual classification are compared in combination with other parameters to analyze and remove bias data. The other parameters include the time of low-clutter antenna injection power and phase change, and the time of the plasma current steady-state stage.

5. The method for calculating the arcing position and time of a low-clutter antenna by using machine learning according to claim 1, characterized in that: The deep convolutional neural network in step 4 uses a large amount of data on different arc starting states that have been identified as input; the different arc starting states include one, two or three of the three arc starting states.

6. The method for calculating the arcing position and time of a low-clutter antenna using machine learning according to claim 1, characterized in that: In step 5, the arc duration at a certain arcing position counted by the machine learning-DBSCAN algorithm is calculated by the collection of frames in which the arc appears discontinuously during the entire discharge process.

Images