Superconducting magnet quench fault automatic discrimination method

Abstract

The invention relates to the technical field of fault discrimination, and discloses a superconducting magnet quench fault automatic discrimination method. The method comprises the following steps: acquiring physical data of the superconducting magnet by using a sensor; performing sliding window framing processing on the acquired physical data to obtain a processed signal frame; performing principal component analysis (PCA) feature extraction on the processed signal frame; and according to the extracted features, training support vector data to describe an SVDD hyper-spherical surface so as to automatically discriminate the quenching fault of the superconducting magnet. Therefore, automatic discrimination of the quench fault of the superconducting magnet can be realized.

Description

Technical Field

[0001] The invention relates to the technical field of fault identification, and in particular to a method for automatically identifying a superconducting magnet quench fault. Background Art

[0002] The unique zero resistance characteristic of superconducting magnets can reduce the AC loss of equipment and effectively reduce its energy consumption. In recent years, with the continuous deepening of research, superconducting magnets have become increasingly valuable in application and research in various fields such as energy upgrades and high-tech such as nuclear magnetic resonance, maglev trains, and magnetic confinement nuclear fusion.

[0003] Superconducting magnets need to work in extreme environments to trigger their superconducting state. Therefore, after being affected by instability, they can easily recover from the superconducting state of zero resistance to the normal conductive state. This phenomenon is called "quenching". The large amount of heat generated when a superconducting magnet quenches can cause damage to its internal structure and local component failures, resulting in a large amount of property losses and even casualties in severe cases, which greatly limits its popularity and application in the industrial field. Therefore, rapid detection of superconducting magnet quench failures has become a basic problem facing the current research field of superconducting materials.

[0004] At present, the machine learning method is mainly used to identify the superconducting magnet quench fault by collecting a large amount of data for classifier or neural network training. For example, acoustic sensors are used to collect signals and extract data features, and superconducting magnet quench faults are detected by dynamically training the automatic encoder network.

[0005] However, the method of training a classifier to identify superconducting magnet quench faults based on collected sensor data requires a massive amount of sample data when training the classifier. For superconducting magnets, it is difficult to obtain a large-scale sample set with high-quality annotations. When the samples are insufficient, the performance of the classifier on unknown data may be reduced, affecting the generalization ability; at the same time, due to the small number of quench faults, its samples are more difficult to obtain, and the characterization of quench fault characteristics is prone to over-reliance on a limited number of fault samples, resulting in overfitting problems. How to identify superconducting magnet quench faults only through normal samples needs to be solved urgently. Summary of the invention

[0006] The invention provides a method for automatically distinguishing a superconducting magnet quench fault, which can solve the problems in the prior art.

[0007] The present invention provides a method for automatically distinguishing a superconducting magnet quench fault, wherein the method comprises:

[0008] Using sensors to collect physical data of superconducting magnets;

[0009] Perform sliding window frame processing on the collected physical data to obtain processed signal frames;

[0010] Perform principal component analysis (PCA) feature extraction on the processed signal frame;

[0011] The support vector data description SVDD hypersphere is trained according to the extracted features to automatically identify the superconducting magnet quench fault.

[0012] Preferably, the sliding window framing process is performed by the following formula to obtain the processed signal frame:

[0013] x ij (n) = φ i (n)·X j (n);

[0014] Among them, X j (n) is the jth physical data collected by the superconducting magnet, φ i (n) is the i-th window function, · is the dot multiplication operation, x ij (n) is the i-th signal frame of the j-th collected physical data obtained after the frame operation.

[0015] Preferably, the window function is a Hamming window, the length of the window function is 0.15 s, and the overlap rate is 70%.

[0016] Preferably, performing principal component analysis (PCA) feature extraction on the processed signal frame includes:

[0017] Calculate signal frame x ij (n) The mean vector μ of the samples in the sample set ij ;

[0018] According to the mean vector μ ij For signal frame x ij (n) De-meaning processing to obtain the data matrix

[0019] Constructing the data matrix The covariance matrix V of

[0020] Perform eigendecomposition on the covariance matrix V and solve the eigenvalue λ ij and the corresponding eigenvector ω ij ;

[0021] According to the eigenvalue λ ij The principal component contribution rates are calculated, and the first d eigenvectors whose sum of principal component contribution rates is greater than 70% are determined as the extracted features.

[0022] Preferably, the signal frame x is calculated by the following formula ij The mean vector μ of the samples in (n) ij :

[0023]

[0024] Where n represents the signal frame x ij (n) is the number of sample points.

[0025] Preferably, the data matrix is ​​obtained by the following formula

[0026]

[0027] Preferably, the data matrix is ​​constructed by The covariance matrix V of is:

[0028]

[0029] Preferably, the sum of the principal component contributions is calculated by the following formula:

[0030]

[0031] Among them, a is the sum of the main component contribution rates, and J is the eigenvalue λ ij The number of

[0032] Preferably, training the support vector data description SVDD hypersphere according to the extracted features to automatically identify the superconducting magnet quench fault includes:

[0033] Calculate signal frame x ij (n) Sample weight of normal samples in the sample set;

[0034] The hypersphere is trained according to the sample weights and the extracted features to obtain the hypersphere related parameters;

[0035] The quench fault of superconducting magnet is automatically identified based on the relevant parameters of the supersphere and the loss function.

[0036] Preferably, the normal sample set includes concentrated samples and outlier samples, and the sample weight of the concentrated samples is greater than the sample weight of the outlier samples.

[0037] Through the above technical solution, the collected physical data can be processed by sliding window framing to obtain the processed signal frame, and then the principal component analysis PCA can be used to extract multi-dimensional physical features. According to the principal component contribution rate, while reducing the redundancy between the original features, the features of appropriate dimensions are selected to retain the state information of the superconducting magnet to the greatest extent. Then, the features extracted by PCA are used to train the SVDD classifier to realize the automatic identification of the superconducting magnet quench fault. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The included drawings are used to provide a further understanding of the embodiments of the present invention, which constitute a part of the specification, are used to illustrate the embodiments of the present invention, and together with the text description, explain the principles of the present invention. Obviously, the drawings in the following description are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0039] Figure 1 A flow chart of a method for automatically distinguishing a superconducting magnet quench fault according to an embodiment of the present invention is shown;

[0040] Figure 2 A schematic diagram showing the increase of the principal component contribution rate with the number of principal components according to an embodiment of the present invention is shown;

[0041] Figure 3 A schematic diagram of the distribution of the first two-dimensional principal components and the classification hyperplane according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0042] It should be noted that, in the absence of conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its application or use. Based on the embodiments in 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.

[0043] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0044] Unless otherwise specifically stated, the relative arrangement of the parts and steps described in these embodiments, numerical expressions and numerical values ​​do not limit the scope of the present invention. At the same time, it should be understood that, for ease of description, the sizes of the various parts shown in the accompanying drawings are not drawn according to the actual proportional relationship. The technology, method and equipment known to ordinary technicians in the relevant field may not be discussed in detail, but in appropriate cases, the technology, method and equipment should be regarded as a part of the authorization specification. In all examples shown and discussed here, any specific value should be interpreted as being merely exemplary, rather than as a limitation. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once a certain item is defined in an accompanying drawing, it does not need to be further discussed in subsequent drawings.

[0045] Figure 1 A flow chart of a method for automatically distinguishing a superconducting magnet quench fault according to an embodiment of the present invention is shown.

[0046] The invention relates to automatic identification of superconducting magnet quench fault based on support vector data description (SVDD) facing sample loss.

[0047] like Figure 1 As shown, an embodiment of the present invention provides a method for automatically distinguishing a superconducting magnet quench fault, wherein the method comprises:

[0048] Using sensors to collect physical data of superconducting magnets (collection of original physical data of superconducting magnets);

[0049] Perform sliding window framing processing on the collected physical data to obtain processed signal frames (window framing);

[0050] Perform principal component analysis (PCA) feature extraction on the processed signal frame (PCA feature extraction);

[0051] The support vector data description (Principal Component Analysis, SVDD) hypersphere (SVDD single classifier training) is trained according to the extracted features to automatically identify the superconducting magnet quench fault.

[0052] Through the above technical solution, the collected physical data can be processed by sliding window framing to obtain the processed signal frame, and then the principal component analysis PCA can be used to extract multi-dimensional physical features. According to the principal component contribution rate, while reducing the redundancy between the original features, the features of appropriate dimensions are selected to retain the state information of the superconducting magnet to the greatest extent. Then, the features extracted by PCA are used to train the SVDD classifier to realize the automatic identification of the superconducting magnet quench fault.

[0053] According to an embodiment of the present invention, sliding window framing processing is performed by the following formula to obtain a processed signal frame:

[0054] x ij (n) = φ i (n)·X j (n);

[0055] Among them, X j (n) is the jth physical data collected by the superconducting magnet (i.e., the original collected physical signal), φ i (n) is the i-th window function, · is the dot multiplication operation, x ij (n) is the i-th signal frame of the j-th collected physical data obtained after the frame division operation (ie, the processed signal frame).

[0056] By framing, the signal can be kept stable in a shorter time while containing enough vibration cycles; the sliding window operation can prevent frequency leakage.

[0057] According to an embodiment of the present invention, the window function is a Hamming window, the length of the window function is 0.15 s, and the overlap rate is 70%.

[0058] According to an embodiment of the present invention, performing principal component analysis (PCA) feature extraction on the processed signal frame includes:

[0059] Calculate signal frame x ij (n) The mean vector μ of the samples in the sample set ij ;

[0060] According to the mean vector μ ij For signal frame x ij (n) De-meaning processing to obtain the data matrix

[0061] Constructing the data matrix The covariance matrix V of

[0062] Perform eigendecomposition on the covariance matrix V and solve the eigenvalue λ ij and the corresponding eigenvector ω ij ;

[0063] According to the eigenvalue λij The principal component contribution rates are calculated, and the first d eigenvectors whose sum of principal component contribution rates is greater than 70% are determined as the extracted features.

[0064] Among them, PCA can be used to express the information of the original data using fewer feature dimensions in the new feature space.

[0065] According to an embodiment of the present invention, the signal frame x is calculated by the following formula: ij The mean vector μ of the samples in (n) ij :

[0066]

[0067] Where n represents the signal frame x ij (n) is the number of sample points.

[0068] According to one embodiment of the present invention, the data matrix is ​​obtained by the following formula:

[0069]

[0070] According to one embodiment of the present invention, the data matrix is ​​constructed by the following formula: The covariance matrix V of is:

[0071]

[0072] According to an embodiment of the present invention, the sum of the principal component contribution rates is calculated by the following formula:

[0073]

[0074] Among them, a is the sum of the main component contribution rates, and J is the eigenvalue λ ij The number of

[0075] According to an embodiment of the present invention, training the support vector data to describe the SVDD hypersphere according to the extracted features to automatically identify the superconducting magnet quench fault includes:

[0076] Calculate signal frame x ij (n) Sample weight of normal samples in the sample set;

[0077] The hypersphere is trained according to the sample weights and the extracted features to obtain the hypersphere related parameters;

[0078] The superconducting magnet quench fault is automatically distinguished according to the hypersphere related parameters and the loss function (discrimination function) (ie, the sample state judgment is realized).

[0079] Among them, SVDD can establish a tight boundary to describe normal samples and thus detect abnormal samples.

[0080] According to an embodiment of the present invention, the normal sample set includes concentrated samples and outlier samples, and the sample weight of the concentrated samples is greater than the sample weight of the outlier samples.

[0081] In other words, we can assign larger weights to “concentrated samples” (high-quality samples) and smaller weights to “outlier samples” (low-quality samples) to achieve a balance in sample value.

[0082] According to an embodiment of the present invention, the sample weight s of a normal sample can be calculated by the following formula: i :

[0083]

[0084] Among them, d i For sample x i (i.e., the eigenvector ω ij ) to the sample center C'. The calculation formula of the sample center C' is:

[0085] According to an embodiment of the present invention, the hypersphere related parameters can be obtained by the following formula:

[0086]

[0087] st||F(x i )-c|| 2 ≤R 2 +ξ i ,

[0088] Where R is the radius of the hypersphere, c is the center of the hypersphere, C is the penalty factor, ξ i is the slack variable, F(x i ) is a nonlinear mapping.

[0089] The above two equations are constrained together to find the function f(R,c,ξ) under the condition st. i ) takes the minimum value, which is the hypersphere related parameter.

[0090] The hypersphere related parameters include the radius of the hypersphere and the center of the hypersphere.

[0091] According to one embodiment of the present invention, the loss function is as follows:

[0092] f(x)=sign(||F(x) i )-c|| 2 -R 2 ),

[0093] Among them, f(x) is the loss function.

[0094] The automatic fault identification method of the present invention is described below with reference to examples.

[0095] Take the physical characteristic sequence of superconducting magnets under normal conditions collected in a superconducting magnet test as an example, including 40 physical quantities such as voltage characteristics and magnetic field characteristics at different measurement points. First, PCA is performed on the 40-dimensional features, and the contribution rate of the principal components changes as follows: Figure 2 As shown in the figure, it can be seen that as the feature dimension increases, the principal component contribution rate gradually increases. When it reaches 30-dimensional features, the principal component contribution rate is close to 100%. Among them, the contribution rate of the first 5 dimensions is as high as 75%, which can effectively reduce the feature dimension while maintaining most of the original information. Then, the SVDD hypersphere is trained based on the PCA extracted features. During the training process, different weight coefficients are assigned according to the sample value to optimize the training process. The distribution of normal samples of superconducting magnets in the two-dimensional feature space and the hypersphere obtained by training are shown in Figure 1. Figure 3 As shown in the figure, the circular sample points represent normal samples of superconducting magnets, and the square sample points represent quench fault samples. It can be seen that the normal samples are mapped to the inside of the hypersphere, and the abnormal samples are excluded from the outside of the hypersphere.

[0096] According to the above analysis, it can be seen that the automatic superconducting magnet quench fault identification method based on support vector data description for sample missing described in the present invention can accurately identify the superconducting magnet quench fault by only using normal samples to train the classification model in the absence of fault samples, thereby meeting the needs of superconducting magnet detection in actual situations.

[0097] It can be seen from the above embodiments that the method of the present invention can be used to distinguish the superconducting magnet quench fault with the missing fault sample: first, PCA is used to extract the multi-dimensional physical features of the superconducting magnet sensor according to the principal component contribution rate, and the state information of the superconducting magnet is retained to the greatest extent while reducing the redundancy of the original data. Then, the SVDD hypersphere is trained according to the PCA extracted features to automatically distinguish the superconducting magnet quench fault. In the training process, the weight coefficient is introduced to give different weights to different samples, wherein the "high-quality" sample is given a larger weight, and the "low-quality" sample is given a smaller weight. After the weight coefficient is introduced, the classification hypersphere will be mainly constrained by the "high-quality" sample, and the influence of a few "low-quality" samples on the hyperplane will be weakened. Using the method of the present invention, the superconducting magnet quench fault can be distinguished only by normal sample training classification hypersphere, and the sample weight is configured by sample quality, and the classifier training process is optimized, so as to achieve better classification effect in the case of missing fault samples, and meet the actual requirements for automatic discrimination of superconducting magnet quench fault.

[0098] In the description of the present invention, it is necessary to understand that the directions or positional relationships indicated by directional words such as "front, back, up, down, left, right", "lateral, vertical, perpendicular, horizontal" and "top, bottom" are usually based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description. Unless otherwise specified, these directional words do not indicate or imply that the devices or elements referred to must have a specific direction or be constructed and operated in a specific direction. Therefore, they cannot be understood as limiting the scope of protection of the present invention. The directional words "inside and outside" refer to the inside and outside relative to the contours of each component itself.

[0099] For ease of description, spatially relative terms such as "above", "above", "on the upper surface of", "above", etc. may be used here to describe the spatial positional relationship between a device or feature and other devices or features as shown in the figure. It should be understood that spatially relative terms are intended to include different orientations of the device in use or operation in addition to the orientation described in the figure. For example, if the device in the accompanying drawings is inverted, the device described as "above other devices or structures" or "above other devices or structures" will be positioned as "below other devices or structures" or "below other devices or structures". Thus, the exemplary term "above" can include both "above" and "below". The device can also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatially relative descriptions used here are interpreted accordingly.

[0100] In addition, it should be noted that the use of terms such as "first" and "second" to limit components is only for the convenience of distinguishing the corresponding components. If not otherwise stated, the above terms have no special meaning and therefore cannot be understood as limiting the scope of protection of the present invention.

[0101] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, 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 shall be included in the protection scope of the present invention.

Claims

1. A method for automatically distinguishing superconducting magnet quench faults. It is characterized in that The method includes: Using sensors to collect physical data of superconducting magnets; Perform sliding window frame processing on the collected physical data to obtain processed signal frames; Perform principal component analysis (PCA) feature extraction on the processed signal frame; The support vector data description SVDD hypersphere is trained according to the extracted features to automatically identify the superconducting magnet quench fault.

2. The method according to claim 1, It is characterized in that The sliding window framing process is performed through the following formula to obtain the processed signal frame: x ij (n)=φ i (n)·X j (n): Among them, X j (n) is the jth physical data collected by the superconducting magnet, φ i (n) is the i-th window function, · is the dot multiplication operation, x ij (n) is the i-th signal frame of the j-th collected physical data obtained after the frame operation.

3. The method according to claim 2, It is characterized in that The window function is a Hamming window, the length of the window function is 0.15s, and the overlap rate is 70%.

4. The method according to claim 3, It is characterized in that The principal component analysis PCA feature extraction of the processed signal frame includes: Calculate signal frame x ij (n) The mean vector μ of the samples in the sample set ij ; According to the mean vector μ ij For signal frame x ij (n) De-meaning processing to obtain the data matrix Constructing the data matrix The covariance matrix V of Perform eigendecomposition on the covariance matrix V and solve the eigenvalue λ ij and the corresponding eigenvector ω ij ; According to the eigenvalue λ ij The principal component contribution rates are calculated, and the first d eigenvectors whose sum of principal component contribution rates is greater than 70% are determined as the extracted features.

5. The method according to claim 4, It is characterized in that The signal frame x is calculated by ij The mean vector μ of the samples in (n) ij : Where n represents the signal frame x ij (n) is the number of sample points.

6. The method according to claim 5, It is characterized in that The data matrix is ​​obtained by the following formula 7. The method according to claim 6, It is characterized in that The data matrix is ​​constructed by The covariance matrix V of is:

8. The method according to claim 7, It is characterized in that The sum of the principal component contributions is calculated using the following formula: Among them, a is the sum of the main component contribution rates, and J is the eigenvalue λ ij The number of 9. The method according to claim 8, It is characterized in that The support vector data description SVDD hypersphere is trained based on the extracted features to automatically identify the superconducting magnet quench fault, including: Calculate signal frame x ij (n) Sample weight of normal samples in the sample set; The hypersphere is trained according to the sample weights and the extracted features to obtain the hypersphere related parameters; The quench fault of superconducting magnet is automatically identified based on the relevant parameters of the supersphere and the loss function.

10. The method according to claim 9, It is characterized in that The normal sample set includes concentrated samples and outlier samples, and the sample weight of concentrated samples is greater than the sample weight of outlier samples.

Images