Spacecraft fault positioning method and device based on knowledge graph
By building the spacecraft knowledge graph and generating transformation matrix methods, the problem of difficulty in accurately locate the cause of the fault in spacecraft fault diagnosis is solved, and more accurate fault location is achieved.
Patent Information
- Application Number
- CN202411569089.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-05
- Publication Date
- 2025-02-14
AI Technical Summary
The prior art is difficult to accurately locate the cause of the fault in spacecraft fault diagnosis, mainly because of the different degree of sensitivity of the fault propagation and different parameters to the fault.
Using a knowledge graph-based method, the knowledge graph of spacecraft historical telemetry data is constructed to characterize the causal relationship between telemetry data, and a transformation matrix is generated by normal telemetry data to perform fault diagnosis of online telemetry data. Then, the candidate data set related to the abnormal data is determined based on the knowledge graph, and the correlation coefficient between each data and the abnormal data is calculated to perform fault location.
Accurate positioning of spacecraft failures is achieved, and the causal relationship between data is characterized by the knowledge graph, which can more accurately identify the real cause of the failure and improve the accuracy of fault location.
Smart Images
Figure CN119443241A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of fault diagnosis technology, and in particular to a method and device for locating a spacecraft fault based on a knowledge graph. Background Art
[0002] Spacecraft have numerous components, complex structures, and thousands of telemetry parameters. During the on-orbit operation of spacecraft, a large amount of data can be used for fault location. At present, in data-driven fault diagnosis, more attention is paid to fault detection. Although fault location can determine that a certain parameter is abnormal, due to the different fault propagation and the different sensitivity of different parameters to the fault, the abnormal parameter is not necessarily the real cause of the fault, resulting in the inability to accurately locate the fault.
[0003] Therefore, there is an urgent need for a spacecraft fault location method and device based on knowledge graph to solve the above problems. Summary of the invention
[0004] The present invention provides a method and device for locating spacecraft faults based on knowledge graphs, which can determine the relationship between telemetry data and abnormal data, and provide a basis for accurate location of spacecraft faults. The technical solution is as follows:
[0005] On the one hand, a method for locating spacecraft faults based on a knowledge graph is provided, the method comprising:
[0006] Constructing a knowledge graph based on the historical telemetry data of the spacecraft; the knowledge graph uses the telemetry data of the spacecraft as entity variables and the causal relationship between the telemetry data as graph relationships;
[0007] generating a transformation matrix based on normal telemetry data of the spacecraft within a preset time period;
[0008] For the online telemetry data at any time, fault diagnosis is performed on the online telemetry data based on the transformation matrix to determine abnormal data in the online telemetry data;
[0009] Based on the knowledge graph, determining a candidate data set related to the abnormal data from the online telemetry data;
[0010] A correlation coefficient between each data in the candidate data set and the abnormal data is calculated to perform fault diagnosis on the spacecraft based on the calculated correlation coefficient.
[0011] On the other hand, a spacecraft fault location device based on a knowledge graph is provided, the device comprising:
[0012] A construction unit, configured to construct a knowledge graph based on historical telemetry data of a spacecraft; the knowledge graph uses the telemetry data of the spacecraft as entity variables and uses the causal relationship between the telemetry data as a graph relationship;
[0013] A generating unit, configured to generate a transformation matrix based on normal telemetry data of the spacecraft within a preset time period;
[0014] A diagnosis unit, configured to perform fault diagnosis on the online telemetry data at any time based on the transformation matrix to determine abnormal data in the online telemetry data;
[0015] a determination unit, configured to determine, from the online telemetry data, a candidate data set related to the abnormal data based on the knowledge graph;
[0016] A calculation unit is used to calculate the correlation coefficient between each data in the candidate data set and the abnormal data, so as to perform fault diagnosis on the spacecraft based on the calculated correlation coefficient.
[0017] On the other hand, a computer-readable storage medium is provided, wherein a computer program is stored in the storage medium, and when the computer program is executed by a processor, the steps of the above-mentioned knowledge graph-based spacecraft fault location method are implemented.
[0018] On the other hand, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the steps of the above-mentioned knowledge graph-based spacecraft fault location method.
[0019] The embodiment of the present invention provides a method for locating spacecraft faults based on a knowledge graph. First, a knowledge graph is constructed using historical telemetry data to characterize the causal relationship between the data through the knowledge graph. Then, a transformation matrix is constructed using normal telemetry data in a certain period of time, so as to diagnose abnormal data in the online telemetry data based on the transformation matrix. Finally, root cause analysis is performed based on the knowledge graph to determine candidate data sets related to the abnormal data and the correlation coefficient between each data in the candidate data set and the abnormal data. The larger the correlation coefficient, the greater the correlation with the fault, thereby providing an important basis for fault location. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0021] Figure 1 It is a flow chart of a method for locating a spacecraft fault based on a knowledge graph provided by an embodiment of the present invention;
[0022] Figure 2 It is a structural diagram of a spacecraft fault location device based on a knowledge graph provided by an embodiment of the present invention;
[0023] Figure 3 It is a hardware architecture diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0024] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, 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 part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0025] The specific implementation of the above concept is described below.
[0026] Please refer to Figure 1 , an embodiment of the present invention provides a method for locating a spacecraft fault based on a knowledge graph, the method comprising:
[0027] Step 100, constructing a knowledge graph based on the historical telemetry data of the spacecraft; the knowledge graph uses the telemetry data of the spacecraft as entity variables and the causal relationship between the telemetry data as the graph relationship;
[0028] Step 102, generating a transformation matrix based on normal telemetry data of the spacecraft within a preset time period;
[0029] Step 104, for the online telemetry data at any time, fault diagnosis is performed on the online telemetry data based on the transformation matrix to determine abnormal data in the online telemetry data;
[0030] Step 106, determining candidate data sets related to abnormal data from online telemetry data based on the knowledge graph;
[0031] Step 108 , calculating the correlation coefficient between each data in the candidate data set and the abnormal data, so as to perform fault diagnosis on the spacecraft based on the calculated correlation coefficient.
[0032] In this embodiment, the knowledge graph is first constructed using historical telemetry data to characterize the causal relationship between the data through the knowledge graph. Then, the transformation matrix is constructed using the normal telemetry data of a certain period of time, so as to diagnose the abnormal data in the online telemetry data based on the transformation matrix. Finally, the root cause analysis is performed based on the knowledge graph to determine the candidate data sets related to the abnormal data and the correlation coefficient between each data in the candidate data set and the abnormal data. The larger the correlation coefficient, the greater the correlation with the fault, thus providing an important basis for fault location.
[0033] Described below Figure 1 How the various steps are performed.
[0034] First, for step 100, the historical telemetry data is a data sequence of a preset length and a preset dimension, and the data sequence is a stationary time series;
[0035] Build a knowledge graph based on the historical telemetry data of spacecraft, including:
[0036] Take the sub-data of each dimension as a variable and calculate the kernel matrix of each variable;
[0037] Based on the calculated kernel matrix, calculate the relationship attribute values between the variables;
[0038] Based on the calculated relationship attribute value, the causal relationship between the variables is determined; wherein the larger the relationship attribute value, the higher the degree of dependence between the two variables;
[0039] Taking each variable as a graph entity and the determined causal relationship between the variables as a graph relationship, an initial knowledge graph is constructed;
[0040] For the initial knowledge graph, triples are extracted and trained based on the TransE model to obtain the embedding vectors between each entity and each graph relationship to generate the final knowledge graph.
[0041] In this step, the historical telemetry data is assumed to be a data sequence X with a length of N and a preset dimension of d. t , If the data of each dimension is a variable, then we need to analyze any two variables x n,i and x n,j The causal relationship between them, j = 1, 2...d, i ≠ j. The variables are taken as graph entities and the obtained causal relationship is taken as graph relationship to construct the corresponding knowledge graph.
[0042] It should be noted that the premise for causal analysis is that the data series is a stationary time series.
[0043] Usually, the following formula is used to determine whether the historical telemetry data is a stationary time series.
[0044]
[0045] In the formula, Δx i,t For the data sequence x i The first-order difference of , t is time, α and β τ is a constant, β is the trend term coefficient, k is the model order of the autoregressive model, ε t is a random error term.
[0046] If ρ = 1, then the data sequence x i There is a unit root, which is a non-stationary sequence; if ρ < 1, then the data sequence x i There is no unit root and it is a stationary series.
[0047] In some embodiments, the causal relationship analysis based on the Hilbert-Schmidt independence criterion (HSIC) calculates the relationship attribute value between two variables. i =[x i,1 x i,2 … x i,n ] T , output variable x j =[x j,1 x j,2 … x j,n ] T Take this as an example to illustrate.
[0048] (1) Calculate the input variables x separately i The kernel matrix and the output variable x j The kernel matrix in:
[0049]
[0050] In the formula, and They are all parameters of the kernel function, which are generally selected based on expert experience.
[0051] (2) Use the following formula to calculate the input variable x i and the output variable x j The relational attribute value Γ i,j :
[0052]
[0053] Where tr(·) is the trace of the matrix, i.e., the sum of the diagonal elements of the matrix; H i,j ∈R n×nis a centralized matrix, the specific form is I is the identity matrix and 1 is a column vector whose elements are all 1.
[0054] When Γ i,j =0, indicating that the input variable x i and the output variable x j Mutually independent, there is no relationship between the two variables; Γ i,j The larger the value, the higher the output variable x j For the input variable x i The higher the degree of dependence.
[0055] Based on the above principle, the relationship attribute values between the variables in the historical telemetry data can be calculated to determine the causal relationship between the variables. Then, taking each variable as a graph entity and the obtained causal relationship as a graph relationship, the initial knowledge graph of the multivariate causal relationship of the spacecraft is constructed.
[0056] In some implementations, for the initial knowledge graph, triples are extracted and trained based on the TransE model to obtain embedding vectors between each entity and each graph relationship, and generate the final knowledge graph, including:
[0057] Extract all triples in the initial knowledge graph; each triple consists of a head entity, a relation, and a tail entity;
[0058] Randomly assign values to entities and relations in each triple within a preset interval to convert them into multi-dimensional vectors to obtain vectorized triples;
[0059] Based on the preset training goal, determine the score function and loss function of the TransE model; the training goal is that the sum of the head entity and its relationship of any triple is equal to its tail entity; the score function is determined based on the Euclidean distance or Manhattan distance between the sum of the head entity and its relationship and its tail entity of the triple;
[0060] For each round of training, the following steps are performed: randomly extract multiple triplets from the vectorized triplets as positive samples, and randomly replace the head entity or tail entity of each positive sample to generate negative samples; for each positive sample and each negative sample, the score function and loss function are optimized and solved based on the gradient optimization algorithm, and the vector elements between each entity and its relationship are continuously updated based on the training objective until the model converges, and the embedding vectors between each entity and its relationship are obtained to generate the final knowledge graph.
[0061] In this embodiment, a triple is represented in the form of (h, r, t), where h represents the head entity, r represents the relationship, and t represents the tail entity. For each extracted unit group, its entity and relationship are randomly assigned values within a preset interval, such as The value is assigned within the interval of , thereby converting it into an m-dimensional vector, and the embedding dimension m is set according to expert experience.
[0062] In addition, the training target of the TransE model is: h+r=t. The score function of the TransE model is set according to the Euclidean distance or Manhattan distance between h+r and t:
[0063] f(h,r,t)=||h+rt||2
[0064] The loss function is set as follows:
[0065]
[0066] Where S is the set of positive samples, (h, r, t) is the positive sample; (h', r, t') is the negative sample corresponding to the positive sample (h, r, t), and γ represents the interval in the loss function (γ>0).
[0067] The gradient-based optimization algorithm is used to optimize and solve the above model, and the vector elements of entities and relationships are continuously updated until the model converges to satisfy h+r=t, and the m-dimensional embedding vector of entities and relationships is obtained.
[0068] For step 102, it includes:
[0069] Normalize the normal telemetry data to obtain a normalized sample matrix; each row in the sample matrix consists of sub-data of a preset dimension;
[0070] Calculate the covariance matrix of the sample matrix;
[0071] Calculate the eigenvalues and eigenvectors of each column in the covariance matrix;
[0072] According to the order of eigenvalues from large to small, the sub-data corresponding to the previous multiple eigenvalues are taken, and the cumulative contribution rate of the taken sub-data is calculated; it is determined whether the calculated cumulative contribution rate is greater than the preset contribution rate; if not, the number of extracted sub-data is increased in sequence until the cumulative contribution rate of the taken sub-data is greater than the preset contribution rate;
[0073] The matrix formed by the eigenvectors of the sub-data finally taken is determined as the transformation matrix of the normal telemetry data.
[0074] In this step, the normal telemetry data within the preset time period is assumed to be X′=[x′1x′2...x′ d ]∈R n×d , and normalize it using the following formula:
[0075]
[0076] In the formula, x′ irepresents the i-th dimension data of the spacecraft, i=1,2,...,d, represents the normalized data, x′ i,min represents the minimum value of the i-th dimension data of the spacecraft in a limited time period under normal circumstances, x′ i,max It represents the maximum value of the i-th dimension data of the spacecraft within a limited time period under normal circumstances.
[0077] The sample matrix obtained after normalization is (n is the number of observed samples, d is the number of variables), use the following steps to transform the sample matrix Perform feature extraction:
[0078] (1) Calculate the sample matrix The covariance matrix c of is:
[0079]
[0080] (2) Calculate the eigenvalue λ of the covariance matrix c i and the corresponding eigenvector u i ,where i=1,2,...,d.
[0081] (3) Arrange the eigenvalues in descending order and calculate the cumulative contribution rate η(k) of the first k data according to the following formula:
[0082]
[0083] The cumulative contribution rate is used to measure the degree of information preservation of the original normal telemetry data by the acquired data; usually, the preset contribution rate is not less than 85%.
[0084] The following formula is used to calculate the eigenvectors corresponding to the first k larger eigenvalues to form the transformation matrix P k :
[0085] P k =[u1 u2 … u k ], P k ∈R d×k , k<d.
[0086] With respect to step 104, for the online telemetry data at any time, fault diagnosis is performed on the online telemetry data based on the transformation matrix to determine abnormal data in the online telemetry data, including:
[0087] Normalizing the online telemetry data to obtain normalized data; the online telemetry data consists of sub-data of preset dimensions;
[0088] Based on the transformation matrix, calculate the square prediction error of the normalized data;
[0089] Determine whether the square prediction error is greater than a preset threshold; if not, determine that the spacecraft has no fault; if so, determine that the spacecraft has a fault, and calculate the sub-square prediction error of the sub-data in each dimension, and determine the sub-data corresponding to the maximum sub-square prediction error as abnormal data.
[0090] In this step, taking the online data at time t as an example, the online data at time t is normalized to obtain The squared prediction error SPE of the normalized data is calculated based on the following formula:
[0091]
[0092] The calculated SPE is compared with a preset threshold. If it exceeds the preset threshold, the spacecraft is considered to have failed. The preset threshold is determined based on experience.
[0093] When it is determined that the spacecraft has failed, the following formula is used to calculate The sub-squared prediction error SPE of the sub-data in each dimension i :
[0094]
[0095] Take SPE i The maximum value in is determined, and the sub-data corresponding to the maximum value is determined as abnormal data.
[0096] For step 106, based on the knowledge graph, candidate data sets related to abnormal data are determined from the online telemetry data, including:
[0097] Determine the location of abnormal data in the knowledge graph;
[0098] Starting from the entity corresponding to the abnormal data in the knowledge graph, based on the embedding vectors between all entities and graph relations in the knowledge graph, find all paths that satisfy h+r=t; where h represents the head entity, r represents the relationship, and t represents the tail entity;
[0099] The set of data corresponding to entities on all paths that meet the requirements is taken as a candidate data set.
[0100] For step 108, the correlation coefficient between each data in the candidate data set and the abnormal data is calculated, including:
[0101] For each data in the candidate dataset, execute:
[0102] Multiply all the relationship attribute values on the path from the data to the abnormal data to obtain the correlation coefficient between the data and the abnormal data.
[0103] Through the above steps, data related to the abnormal data in the online telemetry data can be obtained, and the correlation coefficient between the related data and the abnormal data can be given, and then the designer can determine the final cause of the fault through analysis on this basis.
[0104] like Figure 2 , Figure 3 As shown, an embodiment of the present invention provides a spacecraft fault location device based on a knowledge graph. The device embodiment can be implemented by software, or by hardware or a combination of software and hardware. From the hardware level, Figure 2 As shown, a hardware architecture diagram of a computing device in which a spacecraft fault location device based on a knowledge graph is provided in an embodiment of the present invention is provided. Figure 2 In addition to the processor, memory, network interface, and non-volatile memory shown in the figure, the computing device in which the device is located in the embodiment may also generally include other hardware, such as a forwarding chip responsible for processing messages, etc. Taking software implementation as an example, Figure 3 As shown, as a device in a logical sense, the CPU of the computing device in which it is located reads the corresponding computer program in the non-volatile memory into the internal memory and runs it.
[0105] Please refer to Figure 3 , an embodiment of the present invention provides a spacecraft fault location device based on a knowledge graph, the device comprising:
[0106] A construction unit 300 is used to construct a knowledge graph based on the historical telemetry data of the spacecraft; the knowledge graph uses the telemetry data of the spacecraft as entity variables and the causal relationship between the telemetry data as graph relationships;
[0107] A generating unit 302, configured to generate a transformation matrix based on normal telemetry data of the spacecraft within a preset time period;
[0108] The diagnosis unit 304 is used to perform fault diagnosis on the online telemetry data at any time based on the transformation matrix to determine abnormal data in the online telemetry data;
[0109] A determination unit 306, configured to determine a candidate data set related to the abnormal data from the online telemetry data based on the knowledge graph;
[0110] The calculation unit 308 is used to calculate the correlation coefficient between each data in the candidate data set and the abnormal data, so as to perform fault diagnosis on the spacecraft based on the calculated correlation coefficient.
[0111] In some embodiments, the historical telemetry data is a data sequence of a preset length and a preset dimension, and the data sequence is a stationary time series;
[0112] The construction unit 300 is used to perform the following operations:
[0113] Build a knowledge graph based on the historical telemetry data of spacecraft, including:
[0114] Take the sub-data of each dimension as a variable and calculate the kernel matrix of each variable;
[0115] Based on the calculated kernel matrix, calculate the relationship attribute values between the variables;
[0116] Based on the calculated relationship attribute value, the causal relationship between the variables is determined; wherein the larger the relationship attribute value, the higher the degree of dependence between the two variables;
[0117] Taking each variable as a graph entity and the determined causal relationship between the variables as a graph relationship, an initial knowledge graph is constructed;
[0118] For the initial knowledge graph, triples are extracted and trained based on the TransE model to obtain the embedding vectors between each entity and each graph relationship to generate the final knowledge graph.
[0119] In some implementations, for the initial knowledge graph, triples are extracted and trained based on the TransE model to obtain embedding vectors between each entity and each graph relationship, and generate the final knowledge graph, including:
[0120] Extract all triples in the initial knowledge graph; each triple consists of a head entity, a relation, and a tail entity;
[0121] Randomly assign values to entities and relations in each triple within a preset interval to convert them into multi-dimensional vectors to obtain vectorized triples;
[0122] Based on the preset training goal, determine the score function and loss function of the TransE model; the training goal is that the sum of the head entity and its relationship of any triple is equal to its tail entity; the score function is determined based on the Euclidean distance or Manhattan distance between the sum of the head entity and its relationship and its tail entity of the triple;
[0123] For each round of training, the following steps are performed: randomly extract multiple triplets from the vectorized triplets as positive samples, and randomly replace the head entity or tail entity of each positive sample to generate negative samples; for each positive sample and each negative sample, the score function and loss function are optimized and solved based on the gradient optimization algorithm, and the vector elements between each entity and its relationship are continuously updated based on the training objective until the model converges, and the embedding vectors between each entity and its relationship are obtained to generate the final knowledge graph.
[0124] In some implementations, the generating unit 302 is configured to perform:
[0125] Normalize the normal telemetry data to obtain a normalized sample matrix; each row in the sample matrix consists of sub-data of a preset dimension;
[0126] Calculate the covariance matrix of the sample matrix;
[0127] Calculate the eigenvalues and eigenvectors of each column in the covariance matrix;
[0128] According to the order of eigenvalues from large to small, the sub-data corresponding to the previous multiple eigenvalues are taken, and the cumulative contribution rate of the taken sub-data is calculated; it is determined whether the calculated cumulative contribution rate is greater than the preset contribution rate; if not, the number of extracted sub-data is increased in sequence until the cumulative contribution rate of the taken sub-data is greater than the preset contribution rate;
[0129] The matrix formed by the eigenvectors of the sub-data finally taken is determined as the transformation matrix of the normal telemetry data.
[0130] In some embodiments, the diagnostic unit 304 is configured to perform:
[0131] Normalizing the online telemetry data to obtain normalized data; the online telemetry data consists of sub-data of preset dimensions;
[0132] Based on the transformation matrix, calculate the square prediction error of the normalized data;
[0133] Determine whether the square prediction error is greater than a preset threshold; if not, determine that the spacecraft has no fault; if so, determine that the spacecraft has a fault, and calculate the sub-square prediction error of the sub-data in each dimension, and determine the sub-data corresponding to the maximum sub-square prediction error as abnormal data.
[0134] In some implementations, the determining unit 306 is configured to perform:
[0135] Determine the location of abnormal data in the knowledge graph;
[0136] Starting from the entity corresponding to the abnormal data in the knowledge graph, based on the embedding vectors between all entities and graph relations in the knowledge graph, find all paths that satisfy h+r=t; where h represents the head entity, r represents the relationship, and t represents the tail entity;
[0137] The set of data corresponding to entities on all paths that meet the requirements is taken as a candidate data set.
[0138] In some implementations, the computing unit 308 is configured to perform:
[0139] For each data in the candidate dataset, execute:
[0140] Multiply all the relationship attribute values on the path from the data to the abnormal data to obtain the correlation coefficient between the data and the abnormal data.
[0141] It should be noted that the spacecraft fault location device based on the knowledge graph provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the spacecraft fault location device based on the knowledge graph provided in the above embodiment and the spacecraft fault location method embodiment based on the knowledge graph belong to the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.
[0142] The embodiment of the present application also provides a computer device, please refer to Figure 3 The computer device includes a processor and a memory, in which at least one instruction, at least one program, a code set or an instruction set is stored, and the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the spacecraft fault location method based on the knowledge graph provided in the above-mentioned method embodiments.
[0143] An embodiment of the present application also provides a computer-readable storage medium, on which is stored at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the knowledge graph-based spacecraft fault location method provided in the above-mentioned method embodiments.
[0144] An embodiment of the present application also provides a computer program product, which includes a computer program. A processor of a computer device reads the computer program from a computer-readable storage medium, and the processor executes the computer program, so that the computer device executes the spacecraft fault location method based on the knowledge graph described in any of the above embodiments.
[0145] For the convenience of description, the above system or device is described by dividing it into various modules or units according to its functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.
[0146] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly contributed to the prior art in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.
[0147] Finally, it should be noted that, in this article, relational terms such as first, second, third and fourth are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.
[0148] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.
Claims
1. A method for locating spacecraft faults based on knowledge graph, characterized in that: The method comprises: Constructing a knowledge graph based on the historical telemetry data of the spacecraft; the knowledge graph uses the telemetry data of the spacecraft as entity variables and the causal relationship between the telemetry data as graph relationships; generating a transformation matrix based on normal telemetry data of the spacecraft within a preset time period; For the online telemetry data at any time, fault diagnosis is performed on the online telemetry data based on the transformation matrix to determine abnormal data in the online telemetry data; Based on the knowledge graph, determining a candidate data set related to the abnormal data from the online telemetry data; A correlation coefficient between each data in the candidate data set and the abnormal data is calculated to perform fault diagnosis on the spacecraft based on the calculated correlation coefficient.
2. The method according to claim 1, characterized in that The historical telemetry data is a data sequence of preset length and preset dimension, and the data sequence is a stationary time series; The knowledge graph is constructed based on the historical telemetry data of the spacecraft, including: The sub-data of each dimension is respectively regarded as a variable, and the kernel matrix of each variable is calculated; Based on the calculated kernel matrix, calculate the relationship attribute values between the variables; Based on the calculated relationship attribute value, the causal relationship between the variables is determined; wherein the larger the relationship attribute value, the higher the degree of dependence between the two variables; Taking each variable as a graph entity and the determined causal relationship between the variables as a graph relationship, an initial knowledge graph is constructed; For the initial knowledge graph, triples are extracted and trained based on the TransE model to obtain the embedding vectors between each entity and each graph relationship, and generate the final knowledge graph.
3. The method according to claim 2, characterized in that The process of extracting triples from the initial knowledge graph and training based on the TransE model to obtain embedding vectors between each entity and each graph relationship and generate the final knowledge graph includes: Extract all triples in the initial knowledge graph; each triple consists of a head entity, a relationship, and a tail entity; Randomly assigning values to entities and relations in each triple within a preset interval to convert them into multi-dimensional vectors to obtain vectorized triples; Based on a preset training goal, the score function and loss function of the TransE model are determined; the training goal is that the sum of the head entity and its relationship of any triple is equal to its tail entity; the score function is determined based on the Euclidean distance or Manhattan distance between the sum of the head entity of the triple and its relationship and its tail entity; For each round of training, the following steps are performed: randomly extracting multiple triplets from the vectorized triplets as positive samples, and randomly replacing the head entity or tail entity of each positive sample to generate a negative sample; for each positive sample and each negative sample, optimizing and solving the score function and the loss function based on a gradient optimization algorithm, and continuously updating the vector elements between each entity and its relationship based on the training objective until the model converges, obtaining the embedding vector between each entity and its relationship, and generating the final knowledge graph.
4. The method according to claim 2, characterized in that: The step of generating a transformation matrix based on normal telemetry data of the spacecraft within a preset time period includes: Normalizing the normal telemetry data to obtain a normalized sample matrix; each row in the sample matrix is composed of sub-data of the preset dimension; Calculating the covariance matrix of the sample matrix; Calculate the eigenvalues and eigenvectors of each column in the covariance matrix; According to the order of the eigenvalues from large to small, the sub-data corresponding to the previous multiple eigenvalues are taken, and the cumulative contribution rate of the taken sub-data is calculated; it is determined whether the calculated cumulative contribution rate is greater than the preset contribution rate; if not, the number of extracted sub-data is increased in sequence until the cumulative contribution rate of the taken sub-data is greater than the preset contribution rate; The matrix formed by the eigenvectors of the sub-data finally taken is determined as the transformation matrix of the normal telemetry data.
5. The method according to claim 4, characterized in that The online telemetry data at any time is subjected to fault diagnosis based on the transformation matrix to determine abnormal data in the online telemetry data, including: Normalizing the online telemetry data to obtain normalized data; the online telemetry data is composed of sub-data of the preset dimension; Based on the transformation matrix, calculating the square prediction error of the normalized data; Determine whether the square prediction error is greater than a preset threshold; if not, determine that the spacecraft has no fault; if so, determine that the spacecraft has a fault, and calculate the sub-square prediction error of the sub-data of each dimension, and determine the sub-data corresponding to the maximum sub-square prediction error as abnormal data.
6. The method according to claim 5, characterized in that The determining, based on the knowledge graph, a candidate data set related to the abnormal data from the online telemetry data includes: Determine the location of the abnormal data in the knowledge graph; Taking the entity corresponding to the abnormal data in the knowledge graph as the starting point, based on the embedding vectors between all entities and graph relations in the knowledge graph, find all paths that satisfy h+r=t; where h represents the head entity, r represents the relation, and t represents the tail entity; The set of data corresponding to entities on all paths that meet the requirements is taken as a candidate data set.
7. The method according to claim 5, characterized in that The calculating the correlation coefficient between each data in the candidate data set and the abnormal data includes: For each data in the candidate data set, execute: All relationship attribute values on the path from the data to the abnormal data are multiplied to obtain the correlation coefficient between the data and the abnormal data.
8. A spacecraft fault location device based on knowledge graph, characterized in that: The device comprises: A construction unit, configured to construct a knowledge graph based on historical telemetry data of a spacecraft; the knowledge graph uses the telemetry data of the spacecraft as entity variables and uses the causal relationship between the telemetry data as a graph relationship; A generating unit, configured to generate a transformation matrix based on normal telemetry data of the spacecraft within a preset time period; A diagnosis unit, configured to perform fault diagnosis on the online telemetry data at any time based on the transformation matrix to determine abnormal data in the online telemetry data; a determination unit, configured to determine, from the online telemetry data, a candidate data set related to the abnormal data based on the knowledge graph; A calculation unit is used to calculate the correlation coefficient between each data in the candidate data set and the abnormal data, so as to perform fault diagnosis on the spacecraft based on the calculated correlation coefficient.
9. A computer device, characterized in that: The computer device includes a memory and a processor, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to implement the steps of any one of the methods described in claims 1-7.
10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 7 are implemented.