A hierarchical clustering-based method for extracting fault alarm codes of SA radar

Abstract

The present application belongs to the field of SA radar fault, and relates to a kind of extraction method of SA radar fault alarm code based on hierarchical clustering, comprising the following steps: S1 calculates the similarity distance between alarm codes;S2 calculates the distance between alarm code clusters;S3 alarm code hierarchical clustering process;S4 extracts key alarm code.The present application solves how to distinguish key alarm code and invalid alarm code for SA radar fault, takes key alarm code and invalid alarm code as two kinds of alarm codes, realizes the extraction of key alarm code by distinguishing the difference between the two kinds of alarm codes when fault occurs, and realizes the distinction between key alarm code and invalid alarm code by clustering alarm codes in the form of tree through hierarchical clustering method.

Description

Technical Field

[0001] This invention relates to the field of weather radar technology, and in particular to a method for extracting SA radar fault alarm codes based on hierarchical clustering. Background Technology

[0002] CINRAD / SA next-generation weather radar is currently the mainstream weather radar model in China. Fault identification and determination have always been technical challenges for radar operation and maintenance. The radar data acquisition subsystem (RDA) has an alarm information output function; when the SA radar malfunctions, it generates corresponding RDA alarm information. However, the current RDA system outputs complex alarm codes, including a large number of invalid alarm codes, which constitute a high proportion of the alarm information. This interferes with the rapid and accurate determination of the radar's true fault and makes it difficult to effectively extract key alarm codes.

[0003] Therefore, it is necessary to provide a method for extracting SA radar fault alarm codes based on hierarchical clustering. Since key alarm codes cannot be effectively extracted, this paper proposes a key alarm code extraction method based on hierarchical clustering, which can quickly and accurately extract key alarm codes related to SA radar faults and reduce the interference of invalid alarm codes on the fault diagnosis process. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0005] In view of the problems existing in the above and / or prior art where the alarm codes output by the RDA system are complicated and difficult to use, the present invention is proposed.

[0006] Therefore, the technical problem to be solved by the present invention is to reduce the interference of invalid alarm codes on the fault diagnosis process in the key alarm codes related to SA radar faults.

[0007] To solve the above technical problems, the present invention provides the following technical solution: a method for extracting SA radar fault alarm codes based on hierarchical clustering, comprising: Step 1: Data collection. First, a large amount of SA radar fault alarm code data needs to be collected, including key alarm codes and invalid alarm codes. The number of times the i-th alarm code and the j-th alarm code appear simultaneously on the fault day is TF. ij The number of times it occurs simultaneously on non-fault days is TN. ij .

[0008] The similarity distance between the i-th alarm code and the j-th alarm code is: S(i,j)=

[0009] min(1 / TN ij ,1 / TF ij )

[0010] The more times two alarm codes appear simultaneously on a fault day, or the more times they appear simultaneously on a non-fault day, the smaller the similarity distance, indicating that the two alarm codes are more closely related in terms of fault performance or normal operation.

[0011] Step Two: Data Preprocessing. Based on historical fault records, dates in the historical data are marked as fault days and non-fault days. A fault day is defined as a day in which a fault occurred, and a non-fault day is a day in which no fault occurred. The collected data is preprocessed, and the inter-cluster distance is defined using the average connectivity method. The distance between cluster A and cluster B is defined as: the distance between clusters N in cluster A... a Alarm codes and N in cluster B b The average distance between alarm codes.

[0012] D(A,B)=Σs(i,j) / (N a *N b ), 1≤i≤N a , 1≤j≤N b

[0013] Step 3: Feature extraction. Extract features such as frequency, amplitude, and phase from the preprocessed data for subsequent cluster analysis.

[0014] Step 4: Hierarchical clustering: Use the hierarchical clustering algorithm to perform cluster analysis on the extracted features, group similar data points into a group, and maximize the differences between different groups; (1) Divide all n alarm codes into n clusters;

[0015] (2) According to D(A,B)=Σs(i,j) / (N a *N b ), 1≤i≤N a , 1≤j≤N b The method is used to calculate the inter-cluster distances between any two of n clusters;

[0016] (3) Aggregate the two clusters with the smallest inter-cluster distance to form n-1 clusters, and recalculate the new inter-cluster distance according to the method in step 2;

[0017] (4) If the number of clusters is 2, end the clustering; otherwise, repeat step 3.

[0018] After steps (1)-(4), hierarchical clustering is performed in a loop, and finally two sets of alarm codes, cluster N1 and cluster N2, are obtained.

[0019] Step 5: Determine the threshold. After obtaining the clustering results, a threshold needs to be determined to distinguish between critical alarm codes and invalid alarm codes. Let T be the total number of historical data dates, and T be the number of times the i-th alarm code appears on the fault date. i .

[0020] The effective fault frequency of the i-th alarm is:

[0021] Fi = T i / T

[0022] Calculate the average effective fault frequency of the two clusters in the clustering results of step four, and denote the number of alarm codes of cluster N1 as Q1 and the number of alarm codes of cluster N2 as Q2.

[0023] F N1 =ΣF i / Q1, 1≤i≤Q1

[0024] F N2 =ΣF j / Q2, 1≤j≤Q2

[0025] If F N1 >F N2 Then, cluster N1 is the critical alarm code cluster, and Q1 alarm codes in it are critical alarm codes;

[0026] If F N2 >F N1 Then, cluster N2 is the critical alarm code cluster, and Q2 alarm codes in it are critical alarm codes;

[0027] Step Six: Validate the Results: Apply the threshold to the new dataset and validate the classification results, calculating different values ​​of the inter-cluster distance using different formulas; the calculation method for the inter-cluster distance between Ci and Cj, where: p is any sample belonging to cluster Ci; q is any sample belonging to cluster Cj; the maximum distance formula is:

[0028] dmax(Ci,Cj)=maxp∈Ci,q∈Cjd(p,q),

[0029] dmax(Ci,Cj) represents the maximum distance between two clusters;

[0030] The minimum distance formula is: dmin(Ci,Cj)=minp∈Ci,q∈Cjd(p,q)

[0031] dmin(Ci,Cj) represents the minimum distance between two clusters;

[0032] The average distance formula between groups Ci and Cj is: davg(Ci,Cj)=ninj1p∈Ci∑q∈Cj∑d(p,q);

[0033] dmean(Ci,Cj) represents the distance between the centroids of two clusters; ni is the number of samples in cluster Ci; ni is the number of samples in cluster Cj; ∑p∈C i∑q∈C jd(p,q) represents the distance from each point in cluster Ci to all points in cluster Cj. Here, each point in Ci corresponds to nj distances, and ni points correspond to ni×nj distances.

[0034] By following the steps above, you can successfully distinguish between critical alarm codes and invalid alarm codes.

[0035] As a preferred embodiment of the SA radar fault alarm code extraction method based on hierarchical clustering described in this invention, the hierarchical clustering algorithm combines weather data for classification, then predicts temperature and irradiance, and uses these two predicted weather factors to classify the predicted time period. After classification, different prediction models are selected for training and prediction of data of different day types. Compared with traditional target detection algorithms, the SA radar hierarchical clustering algorithm has higher reliability, is not affected by external light and weather, and can define a cluster as the largest set of densely connected points. Thus, it can divide areas with sufficiently high density into clusters and can discover clusters of arbitrary shapes in noisy data.

[0036] As a preferred embodiment of the SA radar fault alarm code extraction method based on hierarchical clustering described in this invention, the hierarchical clustering algorithm can obtain several cluster centers and all data under the clusters after classification. Compared with traditional target detection algorithms, the SA radar hierarchical clustering algorithm has higher reliability, is not affected by external lighting or weather, and can define a cluster as the largest set of densely connected points. Thus, it can divide areas with sufficiently high density into clusters and can discover clusters of arbitrary shapes in noisy data.

[0037] As a preferred embodiment of the SA radar fault alarm code extraction method based on hierarchical clustering described in this invention, the hierarchical clustering algorithm may also occur in non-fault operation states. Therefore, key alarm codes and invalid alarm codes are treated as two types of alarm codes. By distinguishing the differences between the two types of alarm codes when a fault occurs, the key alarm code can be extracted.

[0038] As a preferred embodiment of the SA radar fault alarm code extraction method based on hierarchical clustering described in this invention, the hierarchical clustering method can cluster alarm codes in the form of a tree, ultimately forming a set of two clusters of alarm codes, thereby realizing the distinction between critical alarm codes and invalid alarm codes.

[0039] The beneficial effects of this invention are as follows: SA radar fault alarm codes can distinguish between critical alarm codes and invalid alarm codes. By treating critical and invalid alarm codes as two categories of alarm codes and differentiating between them based on whether a fault has occurred, critical alarm codes can be extracted. Hierarchical clustering methods can cluster alarm codes in a tree-like structure, ultimately forming two sets of alarm codes, thereby achieving the distinction between critical and invalid alarm codes. Attached Figure Description

[0040] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] in:

[0042] Figure 1 The diagram illustrates the steps of a hierarchical clustering method for extracting key alarm codes in an SA radar fault alarm code extraction method provided by this invention.

[0043] Figure 2 A diagram illustrating the hierarchical clustering process of an SA radar fault alarm code extraction method based on hierarchical clustering provided by this invention.

[0044] Figure 3 A tree diagram illustrating a method for extracting SA radar fault alarm codes based on hierarchical clustering provided by the present invention;

[0045] Figure 4 The flowchart of the SA radar signal clustering and sorting algorithm is provided by the present invention for a method of extracting SA radar fault alarm codes based on hierarchical clustering.

[0046] Figure 5 The diagram shows the principle of the hierarchical clustering algorithm for the extraction method of SA radar fault alarm codes based on hierarchical clustering provided by this invention. Detailed Implementation

[0047] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0048] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0049] Secondly, the present invention will be described in detail with reference to the schematic diagrams. When detailing the embodiments of the present invention, for ease of explanation, the cross-sectional views illustrating the device structure will be partially enlarged, not according to the usual scale. Furthermore, the schematic diagrams are merely examples and should not limit the scope of protection of the present invention. In addition, actual fabrication should include three-dimensional spatial dimensions of length, width, and depth.

[0050] Furthermore, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that mutually excludes other embodiments.

[0051] Example 1

[0052] Reference Figure 1-5 The first embodiment of the present invention provides a method for extracting SA radar fault alarm codes based on hierarchical clustering, including:

[0053] Step 1: Data Collection: First, a large amount of SA radar fault alarm code data needs to be collected, including critical alarm codes and invalid alarm codes. The number of times the i-th alarm code and the j-th alarm code appear simultaneously on the fault day is TF. ij The number of times it occurs simultaneously on non-fault days is TN. ij .

[0054] The similarity distance between the i-th alarm code and the j-th alarm code is: S(i,j)=

[0055] min(1 / TN ij ,1 / TF ij )

[0056] The more times two alarm codes appear simultaneously on a fault day, or the more times they appear simultaneously on a non-fault day, the smaller the similarity distance, indicating that the two alarm codes are more closely related in terms of fault performance or normal operation.

[0057] Step Two: Data Preprocessing: Based on historical fault records, dates in the historical data are marked as fault days and non-fault days. A fault day is defined as a day in which a fault occurred, and a non-fault day is a day in which no fault occurred. The collected data is preprocessed, and the average connectivity method is used to define the inter-cluster distance. The distance between cluster A and cluster B is defined as: the distance between clusters N in cluster A... a Alarm codes and N in cluster B b The average distance between alarm codes.

[0058] D(A,B)=Σs(i,j) / (N a *N b ), 1≤i≤N a, 1≤j≤N b

[0059] Step 3: Feature extraction: Extract features such as frequency, amplitude, and phase from the preprocessed data for subsequent cluster analysis;

[0060] Step 4: Hierarchical clustering: Use the hierarchical clustering algorithm to perform cluster analysis on the extracted features, group similar data points into a group, and maximize the differences between different groups; (1) Divide all n alarm codes into n clusters;

[0061] (2) According to D(A,B)=Σs(i,j) / (N a *N b ), 1≤i≤N a , 1≤j≤N b The method is used to calculate the inter-cluster distances between any two of n clusters;

[0062] (3) Aggregate the two clusters with the smallest inter-cluster distance to form n-1 clusters, and recalculate the new inter-cluster distance according to the method in step 2;

[0063] (4) If the number of clusters is 2, end the clustering; otherwise, repeat step 3.

[0064] After steps (1)-(4), hierarchical clustering is performed in a loop, and finally two sets of alarm codes, cluster N1 and cluster N2, are obtained.

[0065] Step 5: Determine the threshold: After obtaining the clustering results, a threshold needs to be determined to distinguish between critical alarm codes and invalid alarm codes. Let T be the total number of historical data dates, and T be the number of times the i-th alarm code appears on the fault date. i .

[0066] The effective fault frequency of the i-th alarm is:

[0067] Fi = T i / T

[0068] Calculate the average effective fault frequency of the two clusters in the clustering results of step four, and denote the number of alarm codes of cluster N1 as Q1 and the number of alarm codes of cluster N2 as Q2.

[0069] F N1 =ΣF i / Q1, 1≤i≤Q1

[0070] F N2 =ΣF j / Q2, 1≤j≤Q2

[0071] If F N1 >F N2 Then, cluster N1 is the critical alarm code cluster, and Q1 alarm codes in it are critical alarm codes;

[0072] If F N2 >F N1 Then, cluster N2 is the critical alarm code cluster, and Q2 alarm codes in it are critical alarm codes;

[0073] Step Six: Validate the Results: Apply the threshold to the new dataset and validate the classification results, calculating different values ​​of the inter-cluster distance using different formulas; the calculation method for the inter-cluster distance between Ci and Cj, where: p is any sample belonging to cluster Ci; q is any sample belonging to cluster Cj; the maximum distance formula is:

[0074] dmax(Ci,Cj)=maxp∈Ci,q∈Cjd(p,q),

[0075] dmax(Ci,Cj) represents the maximum distance between two clusters;

[0076] The minimum distance formula is: dmin(Ci,Cj)=minp∈Ci,q∈Cjd(p,q)

[0077] dmin(Ci,Cj) represents the minimum distance between two clusters;

[0078] The average distance formula between groups Ci and Cj is: davg(Ci,Cj)=ninj1p∈Ci∑q∈Cj∑d(p,q);

[0079] dmean(Ci,Cj) represents the distance between the centroids of two clusters; ni is the number of samples in cluster Ci; ni is the number of samples in cluster Cj; ∑p∈C i∑q∈C jd(p,q) represents the distance from each point in cluster Ci to all points in cluster Cj. Here, each point in Ci corresponds to nj distances, and ni points correspond to ni×nj distances.

[0080] Hierarchical clustering steps:

[0081] (1) Principle: Similar clusters are merged step by step based on the inter-cluster distance;

[0082] (2) Step 1: Each sample forms a cluster group, which is called atomic clustering;

[0083] (3) Step 2: Calculate the distance between all clusters. You can use one of the following: minimum distance, maximum distance, centroid distance, or average distance.

[0084] (4) Step 3: Merge the two closest clusters, reducing the number of clusters by 1;

[0085] (5) Step 4: Go to step 2 to calculate the distance between clusters. Step 3 merges the nearest clusters. If the algorithm termination condition is met, then stop clustering; otherwise, keep iterating until they are finally merged into one cluster.

[0086] Termination conditions based on hierarchical clustering (aggregate hierarchical clustering) algorithm.

[0087] Hierarchical clustering methods are generally divided into two types: aggregate hierarchical clustering and partitioning hierarchical clustering. Different hierarchical clustering methods are selected for calculation based on different scenarios and device environments.

[0088] Currently, some of the more representative hierarchical clustering algorithms include: BIRCH algorithm, CURE algorithm, ROCK algorithm, and Chameleon algorithm.

[0089] The following is an explanation of the BIRCH algorithm, one of the hierarchical clustering methods.

[0090] BIRCH (Balanced Iterative Reduction and Clustering using Hierarchical Methods) is a comprehensive hierarchical clustering method. It first stores the dataset in a compact compressed format, then performs clustering directly on the compressed dataset (rather than the original dataset), so its I / O cost is linearly related to the size of the dataset. BIRCH is particularly suitable for large datasets and supports incremental or dynamic clustering. The algorithm can generate good clusters after scanning the dataset once; increasing the number of scans can further improve the clustering quality. It introduces two concepts for summarizing cluster information: Clustering Feature and Clustering Feature Tree (CF Tree). A clustering feature is a triple that provides a summary description of the sub-cluster information of an object. Assume that a sub-cluster has N d-dimensional points or objects {X}. i If the sub-cluster has points (i = 1, 2, ..., N), then the CF of this sub-cluster is defined as a triple: CF = {N, LS, SS}. Here, N is the number of points in the sub-cluster; LS is the linear sum of the N points, i.e., X1 + X2 + ... + X... n The centroid of the reaction cluster is located; ss is the sum of squares of the N data points, i.e. Reflecting the size of the clusters. From a statistical point of view, clustering features are statistical summaries of a given sub-cluster: the 0th, 1st, and 2nd moments of the sub-clusters. It records a key metric for calculating clusters and efficiently utilizing storage, as it summarizes information about the sub-clusters, not the stored objects. A CF tree is a highly balanced tree that stores the clustering features of hierarchical clustering. BIRCH employs a multi-stage clustering technique: a single scan of the dataset produces a basic cluster, and one or more additional scans can further improve the clustering quality. The time complexity of this algorithm is O(n), where n is the number of objects. The BIRCH algorithm exhibits linear scalability with the number of objects and good clustering quality. However, each node of a CF tree can only contain a finite number of entries due to size limitations, and a CF tree node does not always correspond to a natural cluster as perceived by the user. Furthermore, BIRCH does not work well if the clusters are not spherical, as it uses the concept of radius or diameter to control the boundaries of the clusters.

[0091] The steps of the BIRCH algorithm are as follows:

[0092] (1) Scan the dataset, group dense data points into sub-clusters, remove sparse data points as outliers, and use the obtained data to build a CF tree initially stored in memory. The algorithm attempts to make the CF tree reflect the clustering information of the dataset as comprehensively as possible within the memory limit;

[0093] (2) Selective compression. The leaf entries in the initial CF tree are scanned to reconstruct a smaller CF tree, while redundant isolated points are removed and denser subclusters are combined into larger subclusters;

[0094] (3) Use an existing global or semi-global clustering algorithm to cluster all leaf entries that cross different node boundaries in the CF tree;

[0095] (4) Selectively refine the clustering results. Using the cluster centroids generated in (3) as seeds, the data points are reassigned to the nearest seeds to obtain a new cluster set, thereby correcting the inaccuracies and refining better clustering results.

[0096] The similarity distance between alarm codes can also be calculated.

[0097] Calculate the similarity distance between alarm codes: Based on historical fault records, dates in the historical data are marked as fault days and non-fault days. A fault day is defined as a day in which a fault occurred, and a non-fault day is a day in which no fault occurred.

[0098] The number of times the i-th alarm code and the j-th alarm code appear simultaneously on the fault day is TF. ijThe number of times it occurs simultaneously on non-fault days is TN. ij .

[0099] The similarity distance between the i-th alarm code and the j-th alarm code is:

[0100] S(i,j)=min(1 / TN ij ,1 / TF ij )

[0101] The more frequently two alarm codes appear together on fault days, or the more frequently they appear together on non-fault days, the smaller the similarity distance, indicating a closer correlation between the two alarm codes in terms of fault manifestation or normal operation. If the two alarm codes are critical alarm codes, the frequency of their simultaneous appearance on fault days will be higher than the frequency of their simultaneous appearance on non-fault days, i.e., 1 / TN. ij <1 / TF ij Conversely, if the alarm codes are two invalid alarm codes, the number of times they occur simultaneously on non-fault days is higher than the number of times they occur simultaneously on fault days, i.e., 1 / TN. ij >1 / TF ij This similarity measurement method ensures that the similarity distance between critical alarm codes is small, while also satisfying the requirement that the similarity distance between non-critical alarm codes is small. This enables the RDA system to effectively and automatically filter out invalid alarm codes and delete them from the network system, reducing the proportion of invalid alarm code information and improving the success rate of radar in quickly and accurately determining the true fault.

[0102] Example 2

[0103] Reference Figure 2 The first embodiment of the present invention provides a method for extracting SA radar fault alarm codes based on hierarchical clustering, including:

[0104] The hierarchical clustering algorithm, after classification, yields several cluster centers and all data within each cluster. It does not require specifying the number of clusters and is insensitive to the choice of distance metric. When the underlying data has a hierarchical structure, the hierarchical structure can be restored. Alarm codes are then distinguished and utilized based on the required alarm codes during operation. Hierarchical clustering can also occur in non-faulty operating states; therefore, critical alarm codes and invalid alarm codes are treated as two categories. By distinguishing the differences between these two categories when a fault has occurred, critical alarm codes are extracted. The hierarchical clustering method can cluster alarm codes in a tree structure, ultimately forming two sets of alarm codes, thereby achieving the distinction between critical and invalid alarm codes.

[0105] First, the distance between alarm code clusters is calculated to distinguish the purpose of the alarm codes. The average connectivity method is generally used to define the inter-cluster distance. The distance between cluster A and cluster B is defined as: the distance between N clusters in cluster A... a Alarm codes and N in cluster B bThe average distance between alarm codes.

[0106] D(A,B)=Σs(i,j) / (N a *N b ), 1≤i≤N a , 1≤j≤N b

[0107] Hierarchical clustering is performed based on the different alarm codes obtained. The following are the demonstration steps of the hierarchical clustering process for alarm codes.

[0108] (1) Divide all n alarm codes into n clusters;

[0109] (2) Calculate the inter-cluster distances between each pair of n clusters using the method in section 2;

[0110] (3) Aggregate the two clusters with the smallest inter-cluster distance to form n-1 clusters, and recalculate the new inter-cluster distance according to the method in step 2;

[0111] (4) If the number of clusters is 2, end the clustering; otherwise, repeat step 3.

[0112] After steps (1)-(4), hierarchical clustering is performed in a loop, and finally two sets of alarm codes, cluster N1 and cluster N2, are obtained.

[0113] Reference Figure 2 This is an example of a hierarchical clustering process for 5 alarm codes a, b, c, d, e. The labeled D(a,b) indicates that the inter-cluster distance between clusters a and b is minimized in the first layer of clustering, and D(ab,c) indicates that the inter-cluster distance between clusters a and c is minimized in the second layer of clustering.

[0114] Extract key alarm codes based on the calculation results:

[0115] Let T be the total number of historical data dates, and let T be the number of times the i-th alarm code appears on the fault date. i .

[0116] The effective fault frequency of the i-th alarm is:

[0117] Fi = T i / T

[0118] Calculate the average effective fault frequency of the two clusters in the clustering results in step 3, and denote the number of alarm codes in cluster N1 as Q1 and the number of alarm codes in cluster N2 as Q2.

[0119] F N1 =ΣF i / Q1, 1≤i≤Q1

[0120] F N2 =ΣF j / Q2, 1≤j≤Q2

[0121] If F N1 >F N2 Then, cluster N1 is the critical alarm code cluster, and Q1 alarm codes in it are critical alarm codes;

[0122] If F N2 >F N1 Then, cluster N2 is the critical alarm code cluster, and Q2 alarm codes in it are critical alarm codes.

[0123] In summary, this invention does not impose specific limitations on the algorithm. The differentiation of SA radar fault alarm codes can be modified according to different scenarios and applications, as long as the differences between the two types of alarm codes in terms of whether a fault has occurred can be distinguished, thus enabling the extraction of key alarm codes. The hierarchical clustering method can cluster alarm codes in a tree-like structure, ultimately forming two sets of alarm codes, thereby distinguishing between key and invalid alarm codes. This allows for the rapid and accurate extraction of key alarm codes related to SA radar faults and reduces the interference of invalid alarm codes in the fault diagnosis process.

[0124] It is important to note that the constructions and arrangements of this application shown in several different exemplary embodiments are merely illustrative. Although only a few embodiments are described in detail in this disclosure, those who consult this disclosure will readily understand that many modifications are possible (e.g., changes in the size, dimensions, structure, shape, and proportions of various elements, as well as parameter values ​​(e.g., temperature, pressure, etc.), mounting arrangements, use of materials, color, orientation, etc.) without substantially departing from the novel teachings and advantages of the subject matter described in this application). For example, an element shown as integrally formed may be composed of multiple parts or elements, the position of elements may be inverted or otherwise altered, and the nature or number or position of discrete elements may be changed or altered. Therefore, all such modifications are intended to be included within the scope of the invention. The order or sequence of any process or method steps may be changed or rearranged according to alternative embodiments. In the claims, any "device plus function" clause is intended to cover the structure described herein that performs the function, and not only structurally equivalent but also equivalent in structure. Other substitutions, modifications, alterations, and omissions may be made in the design, operation, and arrangement of the exemplary embodiments without departing from the scope of the invention. Therefore, the present invention is not limited to the specific embodiments, but extends to various modifications that still fall within the scope of the appended claims.

[0125] Furthermore, in order to provide a concise description of exemplary embodiments, not all features of actual embodiments (i.e., those features that are not relevant to the currently considered best mode for carrying out the invention, or those features that are not relevant to implementing the invention) may be omitted.

[0126] It should be understood that numerous specific implementation decisions can be made during the development of any practical implementation, such as in any engineering or design project. Such development efforts may be complex and time-consuming, but for those skilled in the art who benefit from this disclosure, the development effort will be a routine work of design, manufacturing, and production without requiring much experimentation.

[0127] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for extracting SA radar fault alarm codes based on hierarchical clustering, characterized in that, Includes the following steps: Step 1: Data Collection. First, it's necessary to collect SA radar fault alarm code data, including critical alarm codes and invalid alarm codes. The number of times the i-th alarm code and the j-th alarm code appear simultaneously on the fault day is TF. ij The number of times it occurs simultaneously on non-fault days is TN. ij； The similarity distance between the i-th alarm code and the j-th alarm code is: S(i,j) = min(1 / TN) ij ,1 / TF ij ) The more times two alarm codes appear simultaneously on a fault day, or the more times they appear simultaneously on a non-fault day, the smaller the similarity distance, indicating that the two alarm codes are more closely related in terms of fault performance or normal operation. Step 2: Data preprocessing. Based on historical fault records, dates in the historical data are marked as fault days and non-fault days. A fault day is defined as a day in which a fault occurred, and a non-fault day is a day in which no fault occurred. The collected data is preprocessed, and the average connectivity method is used to define the inter-cluster distance. The distance between cluster A and cluster B is defined as: the distance between cluster A and cluster B is N... a Alarm codes and N in cluster B b The average distance between alarm codes; D(A,B) = Σs(i,j) / (N a *N b ), 1≤i≤N a ,1≤j≤N b Step 3: Feature extraction. Extract features such as frequency, amplitude, and phase from the preprocessed data for subsequent cluster analysis. Step 4: Hierarchical clustering. Use the hierarchical clustering algorithm to perform cluster analysis on the extracted features, group similar data points into groups, and maximize the differences between different groups. (1) Divide all n alarm codes into n clusters; (2) According to D(A,B) = Σs(i,j) / (N a *N b ), 1≤i≤N a , 1≤j≤N b The method is used to calculate the inter-cluster distances between any two of n clusters; (3) Aggregate the two clusters with the smallest inter-cluster distance to form n-1 clusters, and recalculate the new inter-cluster distance according to the method in section 2; (4) If the number of clusters is 2, end the clustering; otherwise, repeat step 3. After steps (1)-(4), hierarchical clustering is performed cyclically, and finally two sets of alarm codes, cluster N1 and cluster N2, are obtained. Step 5: Determine the threshold. After obtaining the clustering results, a threshold needs to be determined to distinguish between critical alarm codes and invalid alarm codes. Let T be the total number of historical data dates, and T be the number of times the i-th alarm code appears on the fault date. i； The effective fault frequency of the i-th alarm is: Fi = T i / T Calculate the average effective fault frequency of the two clusters in the clustering results of step four, and denote the number of alarm codes in cluster N1 as Q1 and the number of alarm codes in cluster N2 as Q. 2； F N1 = ΣF i / Q1,1≤i≤Q1 F N2 = ΣF j / Q2,1≤j≤Q2 If F N1 >F N2 Then, cluster N1 is the critical alarm code cluster, and Q1 alarm codes in it are critical alarm codes; If F N2 >F N1 Then, cluster N2 is the critical alarm code cluster, and Q2 alarm codes in it are critical alarm codes; Step Six: Validate the Results: Apply the threshold to the new dataset and validate the classification results, calculating different inter-cluster distances using different formulas; the calculation method for the inter-cluster distance between Ci and Cj, where: p is any sample belonging to cluster Ci; q is any sample belonging to cluster Cj; the maximum distance formula is: dmax(In,Cj)=maxp Then, q Cjd(p,q), dmax(Ci,Cj) represents the maximum distance between two clusters; The minimum distance formula is: dmin(Ci,Cj)=minpCi,q Cjd(p,q) dmin(Ci,Cj) represents the minimum distance between two clusters; The formula for the average distance between groups Ci and Cj is: davg(Ci,Cj)=ninj1p Ci∑q Cj∑d(p,q); dmean(Ci,Cj) represents the distance between the centroids of two clusters; ni is the number of samples in cluster Ci; ni is the number of samples in cluster Cj; ∑p C i∑q Cjd(p,q) represents the distance from each point in cluster Ci to all points in cluster Cj. Here, each point in Ci corresponds to nj distances, and ni points correspond to ni×nj distances. By following the steps above, you can successfully distinguish between critical alarm codes and invalid alarm codes.

2. The method for extracting SA radar fault alarm codes based on hierarchical clustering according to claim 1, characterized in that: The hierarchical clustering algorithm combines weather data for classification, then predicts temperature and environment, and uses these two predicted weather factors to classify the prediction period. After classification, different prediction models are selected for training and prediction of data of different day types.

3. The method for extracting SA radar fault alarm codes based on hierarchical clustering according to claim 2, characterized in that: After classification, the hierarchical clustering algorithm obtains several cluster centers and all data under each cluster. When the underlying data has a hierarchical structure, the hierarchical structure can be restored, and the data can be distinguished and utilized according to the alarm codes required during operation.

4. The method for extracting SA radar fault alarm codes based on hierarchical clustering according to claim 3, characterized in that: The hierarchical clustering algorithm may also occur in non-faulty operating states. Therefore, key alarm codes and invalid alarm codes are treated as two types of alarm codes. By distinguishing the differences between the two types of alarm codes when a fault occurs, the key alarm codes can be extracted.

5. The method for extracting SA radar fault alarm codes based on hierarchical clustering according to claim 4, characterized in that: The hierarchical clustering method can cluster alarm codes in a tree structure, ultimately forming two sets of alarm codes, thereby distinguishing between critical alarm codes and invalid alarm codes.

Images