A phased array element fault diagnosis method based on parallel deep learning, an electronic device and a storage medium
By dividing the phased array antenna array into subarrays and using amplitude data for parallel deep learning, faulty units can be diagnosed and located quickly and effectively, solving the problems of complexity and high cost in traditional methods and achieving efficient fault detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- YANGTZE DELTA REGION INST OF UNIV OF ELECTRONICS SCI & TECH OF CHINE (HUZHOU)
- Filing Date
- 2022-12-06
- Publication Date
- 2026-05-29
AI Technical Summary
Traditional phased array element fault diagnosis methods are highly complex in large-scale arrays, and machine learning models become more costly when classifying multiple types of faults, making it difficult to effectively diagnose and locate faulty elements.
By employing a parallel deep learning approach, the phased array antenna array is divided into multiple subarrays. The amplitude data of the radiation pattern is used for initial grouping and training of a deep convolutional neural network to generate a probability prediction vector for fault units, thereby quickly locating the fault location.
It improves the efficiency and accuracy of fault diagnosis, avoids the use of expensive phase data measurements, reduces the complexity of model output, and is suitable for fault detection of large-scale arrays.
Smart Images
Figure CN116027284B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of fault diagnosis technology, and more specifically, to a fault diagnosis method, electronic device, and storage medium for phased array elements based on parallel deep learning. Background Technology
[0002] In recent years, phased array antennas have been widely used in military and civilian fields due to their numerous advantages, and their array sizes have become increasingly larger. Correspondingly, the failure rate of array elements has also increased proportionally. To ensure the performance of the antenna system, it is necessary to effectively diagnose and locate element failures in the array.
[0003] Traditional methods for diagnosing phased array element faults require using test signals to troubleshoot each element individually, a process whose complexity increases with the array size. The radiation pattern of a phased array antenna represents its overall operational characteristics; if changes in this pattern could be used to diagnose antenna faults, diagnostic efficiency would be significantly improved. While amplitude and phase data can be obtained from the radiation pattern, in high-frequency applications, vector measurement equipment is too expensive, and phase angle measurements are difficult in engineering. Therefore, the most feasible option is to use only the amplitude data from the radiation pattern to diagnose element faults.
[0004] Machine learning and deep learning have wide applications in array element fault diagnosis, but they are often treated as single-label classification problems during modeling. In this case, the cost of fault classification increases rapidly with the size of the array. When the number of faulty array elements increases, the output of multi-class machine learning / deep learning models becomes difficult to process. Therefore, using a two-level fault and classification model during modeling, and dividing the antenna array into multiple subarrays in the initial stage, can effectively improve the efficiency and accuracy of fault classification. Summary of the Invention
[0005] In view of this, the purpose of this invention is to provide a method for effectively identifying and locating faulty units by using amplitude data in the radiation pattern of the phased array antenna when an array unit in a phased array radar fails.
[0006] A first aspect of the present invention provides a method for fault diagnosis of phased array elements based on parallel deep learning, the method comprising:
[0007] Obtain the radiation pattern of the phased array antenna array under fault conditions, and generate amplitude data based on the radiation pattern;
[0008] Initialization grouping is performed based on the amplitude data to obtain group training data; the group training data includes a fault-free unit group, a faulty unit group, and multiple faulty unit groups.
[0009] The phased array antenna array is divided into S sub-arrays. For the training data in each sub-array, a deep convolutional neural network model (DCNN) is trained separately to obtain S DCNN models.
[0010] Fault diagnosis is performed on the fault test data using S DCNN models, and a prediction vector containing the fault probability of the corresponding fault unit is output. The location of the predicted fault unit is determined according to the number of fault units and the fault probability is ordered from largest to smallest.
[0011] Preferably, acquiring the radiation pattern under phased array antenna array failure and generating amplitude data based on the radiation pattern includes:
[0012] Based on mathematical modeling Radiation pattern under phased array antenna array failure:
[0013]
[0014] in, For the antenna wavelength, These are the array units in and Distance along the axial direction;
[0015] The excitation current of the cell in row m and column n is assigned the following value:
[0016]
[0017] List the radiation patterns for all failure scenarios, and in Two-dimensional amplitude sampling is performed on the integer interval to obtain a set of dimensions. The amplitude data.
[0018] Preferably, initial grouping is performed based on the amplitude data to obtain grouped training data; the grouped training data includes a fault-free unit group, a faulty unit group, and multiple faulty unit groups, including:
[0019] All amplitude data were compared with amplitude data under fault-free conditions to measure similarity, using the peak signal-to-noise ratio (PSNR) as the metric.
[0020]
[0021] This yields a set of peak signal-to-noise ratios (PSNRs). By setting appropriate thresholds, the PSNRs of all fault scenarios are grouped according to the number of faulty units. The thresholds can be set as follows:
[0022]
[0023] in, It is the average peak signal-to-noise ratio. The standard deviation of the peak signal-to-noise ratio. Take a value from 1, 2, and 3;
[0024] This invention requires experimental adjustment. The value of is obtained when the peak signal-to-noise ratio (PSNR) of a fault-free group in the training data can be distinguished from that of a faulty group. When the peak signal-to-noise ratio of a single faulty unit can be distinguished from the peak signal-to-noise ratio of multiple faulty units, the result is... ;
[0025] When the peak signal-to-noise ratio of an unknown scene When a fault-free unit is determined, it is assigned to the fault-free unit group.
[0026] When the peak signal-to-noise ratio of an unknown scene and At that time: it is determined that there is a faulty unit, and it is assigned to a faulty unit group;
[0027] When the peak signal-to-noise ratio of an unknown scene When: it is determined that there are multiple faulty units, and they are grouped into multiple faulty unit groups.
[0028] Preferably, the phased array antenna array is divided into S subarrays, and the training data in each subarray is used to train a deep convolutional neural network (DCNN) model separately, resulting in S DCNN models, including:
[0029] Will The phased array antenna array is divided into S subarrays. Under each fault scenario, multiple sets of training data are generated according to different signal-to-noise ratios. The multiple sets of training data are classified according to the fault scenario to obtain multiple fault units. Each fault category represents a fault unit.
[0030] Deep convolutional neural networks are trained based on the training data corresponding to each fault category, and feature vectors matching the probabilities of different fault categories are extracted to obtain S DCNN models.
[0031] Preferably, fault diagnosis is performed on the fault test data using S DCNN models, outputting a prediction vector containing the fault probability of the corresponding fault unit; the predicted fault unit location is determined according to the number of fault units and in descending order of fault probability, including:
[0032] Acquire fault test data, calculate its similarity score with amplitude data under fault-free conditions, and determine the number of fault units corresponding to the fault test data;
[0033] Fault diagnosis is performed on the fault test data using S DCNN models, resulting in S vectors containing the probability of the corresponding fault unit. The probabilities of the same fault unit in the S vectors are compared, and only the highest probability is saved. Finally, a prediction vector containing the fault probability of the corresponding fault unit is obtained.
[0034] The fault location is determined based on the number of fault units. Assuming that the fault test data contains 3 fault units as determined by the similarity score, the 3 mapping units with the highest probability are found in the output prediction vector. The location of the unit mapping is the fault location diagnosed by this invention.
[0035] Furthermore, a third aspect of the present invention provides an electronic device comprising: one or more processors, and a memory for storing one or more computer programs; characterized in that the computer programs are configured to be executed by the one or more processors, and the programs include steps for performing the phased array element fault diagnosis method based on parallel deep learning as described in the first aspect above.
[0036] Furthermore, a fourth aspect of the present invention provides a storage medium storing a computer program; the program is loaded and executed by a processor to implement the steps of the phased array element fault diagnosis method based on parallel deep learning as described in the first aspect above.
[0037] In this invention, the radiation pattern of a phased array antenna array under fault conditions is acquired, and amplitude data is generated based on the radiation pattern. Initialization grouping is performed based on the amplitude data to obtain group training data. The group training data includes a group of fault-free units, a group of faulty units, and multiple groups of faulty units. The phased array antenna array is divided into S subarrays. For the training data in each subarray, a deep convolutional neural network (DCNN) model is trained separately to obtain S DCNN models. Fault diagnosis is performed on the fault test data based on the S DCNN models, outputting a prediction vector containing the fault probability of the corresponding faulty unit. The predicted faulty unit location is determined according to the number of faulty units and their fault probabilities from largest to smallest. Compared to existing technologies, by dividing a large array into multiple subarrays for testing, faulty units can be quickly located and the accuracy of location can be effectively improved. A better indicator is introduced to evaluate the impact of faulty units on the radiation pattern, thereby determining whether a fault exists. Only amplitude data is used to diagnose antenna faults, eliminating the need for phase data, which is difficult to acquire. A two-level fault detection and classification model is proposed to avoid affecting the model output efficiency due to the increase in array size. Attached Figure Description
[0038] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0039] Figure 1 This is a flowchart of a phased array element fault diagnosis method based on parallel deep learning disclosed in an embodiment of the present invention;
[0040] Figure 2 This is a schematic diagram of the overall implementation principle disclosed in the embodiments of the present invention;
[0041] Figure 3 This is a schematic diagram of the model training phase disclosed in an embodiment of the present invention. Detailed Implementation
[0042] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art.
[0043] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.
[0044] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.
[0045] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.
[0046] It should be noted that "multiple" as mentioned in this article refers to two or more.
[0047] The implementation details of the technical solutions in the embodiments of this application are described in detail below:
[0048] In this embodiment, a large antenna array is divided into several small antenna subarrays. A set of parallel deep convolutional neural network (DCNNs) models are developed and trained using amplitude data from the radiation pattern of the phased array antenna.
[0049] like Figure 1 The diagram shows a flowchart of a phased array element fault diagnosis method based on parallel deep learning, according to an embodiment of the present invention. The method includes:
[0050] Step S1: Obtain the radiation pattern under phased array antenna array failure, and generate amplitude data based on the radiation pattern.
[0051] Specifically, this step is the preparation stage, which first involves data extraction and sampling, and then obtaining the results based on mathematical modeling. Radiation pattern under phased array antenna array failure:
[0052]
[0053] in, For the antenna wavelength, These are the array units in and Distance along the axis The excitation current of the m-th row and n-th column element can be assigned the following value in this method:
[0054]
[0055] List the radiation patterns under all possible failure scenarios, and in Two-dimensional amplitude sampling is performed on the integer interval to obtain a set of dimensions. The amplitude data.
[0056] Step S2: Perform initial grouping based on the amplitude data to obtain group training data; the group training data includes a fault-free unit group, a faulty unit group, and multiple faulty unit groups.
[0057] Specifically, the data obtained in the preparation phase is initialized and grouped. For example... Figure 2 The diagram shown illustrates the overall implementation principle of this example. Among them, Figure 2 In the data grouping section, all amplitude data are compared with amplitude data under fault-free conditions to measure similarity. The metric used in this method is the peak signal-to-noise ratio (PSNR), defined as:
[0058]
[0059] This yields a set of peak signal-to-noise ratios (PSNRs). By setting an appropriate threshold, the PSNRs of all fault scenarios are grouped according to the number of faulty units. The threshold can be set as follows:
[0060]
[0061] in, It is the average peak signal-to-noise ratio. The standard deviation of the peak signal-to-noise ratio. In most cases, the value is taken from 1, 2, or 3. This method requires extensive experimentation and adjustment to group scenarios according to the number of faults, with the final result being:
[0062] When the peak signal-to-noise ratio of an unknown scene When the fault-free unit is determined, it is assigned to group 1.
[0063] When the peak signal-to-noise ratio of an unknown scene and At that time: it was determined that there was a faulty unit, and it was assigned to group 2.
[0064] When the peak signal-to-noise ratio of an unknown scene At that time: it was determined that there were multiple faulty units and they were assigned to group 3.
[0065] For the sake of brevity, only three groupings are given. In actual engineering, the groupings should distinguish all possible numbers of faults, with a number of... (From no faults to all faults), the pseudocode of the method is given by Algorithm 1.
[0066]
[0067] Step S3: Divide the phased array antenna array into S subarrays, train a deep convolutional neural network model (DCNN) on the training data in each subarray separately, and obtain S DCNN models.
[0068] Among them, The phased array antenna array is divided into S subarrays. Under each fault scenario, multiple sets of training data are generated according to different signal-to-noise ratios. The multiple sets of training data are classified according to the fault scenario to obtain multiple fault units. Each fault category represents a fault unit. Based on the training data corresponding to each fault category, a deep convolutional neural network is trained to extract feature vectors that match the probabilities of different fault categories, thereby obtaining S DCNN models.
[0069] Specifically, this step is the training phase. Among them, such as... Figure 2 The training model part will use the antenna array The array is divided into S subarrays, assuming each subarray contains 4 array units. For each fault scenario, multiple sets of training data are generated based on different signal-to-noise ratios. All training data containing faults in unit 1 are assigned to category 1, all training data containing faults in unit 2 are assigned to category 2, and so on. For the training data in each subarray, a deep convolutional neural network is trained separately, extracting the feature vector for each category, thus obtaining S training models. Further illustration of step S3 is provided by... Figure 3 Provided.
[0070] Furthermore, performance evaluation metrics for model training include correctness, accuracy, recall, and F1 score:
[0071]
[0072]
[0073]
[0074]
[0075] in, Indicates the number of classes. and These represent the number of true positives and the number of false positives, respectively. and These represent true negative numbers and false negative numbers, respectively.
[0076] Step S4: Perform fault diagnosis on the fault test data using S DCNN models and output a prediction vector containing the fault probability of the corresponding fault unit; determine the predicted fault unit location according to the number of fault units and in descending order of fault probability.
[0077] Step S4 specifically includes: acquiring fault test data, calculating its similarity score with amplitude data under fault-free conditions, and determining the number of fault units corresponding to the fault test data;
[0078] Fault diagnosis is performed on the fault test data using S DCNN models, resulting in S vectors containing the probability of the corresponding fault unit. The probabilities of the same fault unit in the S vectors are compared, and only the highest probability is saved. Finally, a prediction vector containing the fault probability of the corresponding fault unit is obtained.
[0079] The fault location is determined based on the number of fault units. Assuming that the fault test data contains 3 fault units as determined by the similarity score, the three mapping units with the highest probability are found in the output prediction vector. The location of the unit mapping is the fault location diagnosed by this invention.
[0080] Specifically, in this embodiment, step S4 is the testing phase. In the testing phase, the similarity between the tested amplitude data samples and the amplitude data samples under normal operation is first measured. By comparing this similarity with a threshold, the number of faulty units is obtained. Then, the tested amplitude data samples are fed into S trained models for testing, resulting in a vector of matching probabilities, representing the probability of the corresponding unit failing.
[0081] like Figure 2 In the test model section, for randomly generated test samples, the similarity score between the sample and the amplitude under fault-free conditions is first calculated (using Algorithm 1) to determine the number of faulty units in the test sample. Then, the S trained models obtained in step two are used to diagnose faults in the test samples, and the output is a prediction vector containing the fault probability of the corresponding unit. Based on the number of faulty units, the specific location of the predicted faulty unit can be obtained by ranking the probabilities from largest to smallest. The pseudocode for the above method is given in Algorithm 2.
[0082]
[0083] In this embodiment, the radiation pattern of a phased array antenna array under fault conditions is acquired, and amplitude data is generated based on the radiation pattern. Initialization grouping is performed based on the amplitude data to obtain group training data. The group training data includes a group of fault-free units, a group of one faulty unit, and multiple groups of faulty units. The phased array antenna array is divided into S subarrays. For the training data in each subarray, a deep convolutional neural network (DCNN) model is trained separately, resulting in S DCNN models. Fault diagnosis is performed on the fault test data based on the S DCNN models, outputting a prediction vector containing the fault probability of the corresponding faulty unit. The predicted faulty unit location is determined according to the number of faulty units and their fault probabilities from largest to smallest. Compared to existing technologies, by dividing the large array into multiple subarrays for testing, faulty units can be quickly located and the accuracy of location can be effectively improved. A better indicator is introduced to evaluate the impact of faulty units on the radiation pattern, thereby determining whether a fault exists. Only amplitude data is used to diagnose antenna faults, eliminating the need for phase data, which is difficult to acquire. A two-level fault detection and classification model is proposed to avoid affecting the model output efficiency due to the increase in array size.
[0084] Furthermore, embodiments of this application also disclose an electronic device comprising: one or more processors, and a memory for storing one or more computer programs; characterized in that the computer programs are configured to be executed by the one or more processors, and the programs include steps for performing the phased array element fault diagnosis method based on parallel deep learning as described in the first aspect above.
[0085] Furthermore, embodiments of this application also provide a storage medium storing a computer program; the program is loaded and executed by a processor to implement the steps of the phased array element fault diagnosis method based on parallel deep learning as described in the first aspect above.
[0086] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0087] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices or units, or may be electrical, mechanical or other forms of connection.
[0088] The units described as separate components may or may not be physically separate. As will be appreciated by those skilled in the art, the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0089] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0090] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or grid device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0091] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for fault diagnosis of phased array elements based on parallel deep learning, characterized in that, The method includes: Obtain the radiation pattern of the phased array antenna array under fault conditions, and generate amplitude data based on the radiation pattern; Initialization grouping is performed based on the amplitude data to obtain group training data; the group training data includes a fault-free unit group, a faulty unit group, and multiple faulty unit groups. The phased array antenna array is divided into S sub-arrays. For the training data in each sub-array, a deep convolutional neural network model (DCNN) is trained separately to obtain S DCNN models. Fault diagnosis is performed on the fault test data using S DCNN models, and a prediction vector containing the fault probability of the corresponding fault unit is output; the location of the predicted fault unit is determined according to the number of fault units and the fault probability is ordered from largest to smallest. Obtain the radiation pattern of the phased array antenna array under fault conditions, and generate amplitude data based on the radiation pattern, including: Based on mathematical modeling Radiation pattern under phased array antenna array failure: in, For the antenna wavelength, These are the array units in and Distance along the axial direction; The excitation current of the cell in row m and column n is assigned the following value: List the radiation patterns for all failure scenarios, and in Two-dimensional amplitude sampling is performed on the integer interval to obtain a set of dimensions. The amplitude data; Initialization grouping is performed based on the amplitude data to obtain group training data; the group training data includes a fault-free unit group, a faulty unit group, and multiple faulty unit groups, including: All amplitude data were compared with amplitude data under fault-free conditions to measure similarity, using the peak signal-to-noise ratio (PSNR) as the metric. This yields a set of peak signal-to-noise ratios (PSNRs). By setting appropriate thresholds, the PSNRs of all fault scenarios are grouped according to the number of faulty units. The thresholds can be set as follows: in, It is the average peak signal-to-noise ratio. The standard deviation of the peak signal-to-noise ratio. Take a value from 1, 2, and 3; Adjusted through experiments The value of is obtained when the peak signal-to-noise ratio (PSNR) of a fault-free group in the training data can be distinguished from that of a faulty group. When the peak signal-to-noise ratio of a single faulty unit can be distinguished from the peak signal-to-noise ratio of multiple faulty units, the result is... ; When the peak signal-to-noise ratio of an unknown scene When a fault-free unit is determined, it is assigned to the fault-free unit group. When the peak signal-to-noise ratio of an unknown scene and At that time: it is determined that there is a faulty unit, and it is assigned to a faulty unit group; When the peak signal-to-noise ratio of an unknown scene When: it is determined that there are multiple faulty units, and they are grouped into multiple faulty unit groups.
2. The phased array element fault diagnosis method based on parallel deep learning according to claim 1, characterized in that, The phased array antenna array is divided into S subarrays. For the training data in each subarray, a deep convolutional neural network (DCNN) model is trained separately, resulting in S DCNN models, including: Will The phased array antenna array is divided into S subarrays. Under each fault scenario, multiple sets of training data are generated according to different signal-to-noise ratios. The multiple sets of training data are classified according to the fault scenario to obtain multiple fault units. Each fault category represents a fault unit. Deep convolutional neural networks are trained based on the training data corresponding to each fault category, and feature vectors matching the probabilities of different fault categories are extracted to obtain S DCNN models.
3. The phased array element fault diagnosis method based on parallel deep learning according to claim 2, characterized in that, Fault diagnosis is performed on the fault test data using S DCNN models, outputting a prediction vector containing the fault probability of the corresponding fault unit; based on the number of fault units, the predicted fault unit locations are determined in descending order of fault probability, including: Acquire fault test data, calculate its similarity score with amplitude data under fault-free conditions, and determine the number of fault units corresponding to the fault test data; Fault diagnosis is performed on the fault test data using S DCNN models, resulting in S vectors containing the probability of the corresponding fault unit. The probabilities of the same fault unit in the S vectors are compared, and only the highest probability is saved. Finally, a prediction vector containing the fault probability of the corresponding fault unit is obtained. The fault location is determined based on the number of faulty units. Assuming that the fault test data contains 3 faulty units as determined by the similarity score, the 3 mapping units with the highest probability are found in the output prediction vector, and the location of the unit mapping is the diagnosed fault location.
4. An electronic device, the electronic device comprising: One or more processors, a memory for storing one or more computer programs; characterized in that the computer programs are configured to be executed by the one or more processors, the programs including steps for performing the phased array element fault diagnosis method based on parallel deep learning as described in any one of claims 1-3.
5. A storage medium storing a computer program; the program being loaded and executed by a processor to implement the steps of the phased array element fault diagnosis method based on parallel deep learning as described in any one of claims 1-3.