Fault detection and isolation method and device for complex equipment of spacecraft and medium

Abstract

The invention discloses a fault detection and isolation method and device for spacecraft complex equipment and a medium, and relates to the field of fault recognition, and the method comprises the steps: inputting the current operation state data of the spacecraft complex equipment into an operation state prediction model for prediction, so as to obtain the future operation state data distribution; the operation state prediction model is obtained by training a variational Kupman anomaly detector by adopting a time sequence training sample set of the spacecraft complex equipment; based on the current operation state data and future operation state data distribution, calculating a prediction error matrix and a corresponding T2 value and SPE statistical magnitude, and then determining the working state of the complex equipment of the spacecraft based on a preset limit value; and when the working state is a fault, calculating a deviation score of each operation state variable and performing fault isolation so as to determine an operation state fault variable. According to the invention, effective fault detection and fault variable screening of the complex equipment of the spacecraft are realized.

Description

Technical Field

[0001] The present application relates to the field of fault identification technology, and in particular to a method, device and medium for fault detection and isolation of complex equipment on a spacecraft. Background Art

[0002] Effective fault detection and isolation (FDI) is crucial for time series data monitoring of complex spacecraft equipment, not only to ensure the safety of personnel and equipment, but also to maintain stable product quality. The main goal of fault detection is to identify abnormal behavior, while fault isolation involves identifying the variables that are most likely to cause failures. However, due to the inherent complexity of time series data of spacecraft equipment, it remains a major challenge to accurately model the system using expert knowledge to directly achieve fault detection and isolation. Therefore, it is crucial to develop advanced automated FDI systems to support operators in managing abnormal situations. In recent years, data mining technology has made significant progress, which makes it possible to extract information from the collected data, thereby facilitating the realization of fault detection and isolation.

[0003] Typical data-driven multivariate statistical methods include principal component analysis, partial least squares regression, and independent component analysis, which are often used for fault detection and isolation of time series data of complex spacecraft equipment. However, these methods assume that process variables are linear and normal, which limits their applicability in nonlinear real process data. In order to address the limitations of traditional methods, deep learning methods have been introduced for fault detection and isolation. These models learn data representations through nonlinear activation functions and complex network architectures, and can effectively handle complex nonlinear relationships. However, these deep learning models do not explicitly model dynamic systems. In addition, due to their complexity, deep learning models are often regarded as "black boxes" and it is difficult to explain their internal processes and decision-making mechanisms. Summary of the invention

[0004] The purpose of this application is to provide a method, device and medium for fault detection and isolation of complex equipment on a spacecraft, thereby achieving effective fault detection and fault variable screening of complex equipment on a spacecraft.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a method for fault detection and isolation of complex equipment of a spacecraft, comprising:

[0007] Acquiring current operating state data of complex equipment of a spacecraft; the current operating state data comprising a plurality of operating state variables;

[0008] The current operating state data is input into the operating state prediction model for prediction to obtain the distribution of future operating state data; wherein the operating state prediction model is obtained by training the variational Koopman anomaly detector using a time series training sample set of complex equipment of the spacecraft; the time series training sample set includes multiple operating state data within a preset time length; the variational Koopman anomaly detector is determined based on a variational autoencoder and a Koopman matrix layer;

[0009] Based on the current operating state data and the future operating state data distribution, the prediction error matrix and the corresponding T are calculated. 2 Value, SPE statistic;

[0010] Based on the preset limit, according to the T corresponding to the current operating status data 2 The working state of the complex equipment of the spacecraft is determined by using the value and the SPE statistic; the working state is a fault or no fault;

[0011] When the working state is a fault, the deviation score of each operating state variable is calculated according to the current operating state data and the future operating state data distribution, and fault isolation is performed to determine the operating state fault variable.

[0012] In a second aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a method for fault detection and isolation of complex equipment of a spacecraft.

[0013] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for fault detection and isolation of complex equipment on a spacecraft.

[0014] According to the specific embodiments provided by the present application, the present application has the following technical effects: the present application provides a method, device and medium for fault detection and isolation of complex equipment of spacecraft, and combines the variational autoencoder and the Koopman operator theory to obtain a variational Koopman anomaly detector, which is trained to predict the distribution of future operating status data. Due to the combination of the variational autoencoder and the Koopman operator theory, it is completely data-driven and does not require expert knowledge in related fields. In addition, the variational Koopman anomaly detector adopted overcomes the dependence of traditional methods on the assumptions of linearity and normality of process variables, as well as the difficulty of deep learning methods in explaining internal processes and decision-making mechanisms. The uncertainty estimate generated by it can be used for fault detection in time series data of complex equipment of spacecraft, and then the deviation score set based on the uncertainty output of the variational Koopman anomaly detector is used to define the degree of deviation of each variable after the fault, so as to intuitively and effectively screen out fault-related variables, and also perform fault isolation. The present application realizes the effective identification of faults and fault isolation variable screening in time series data of complex equipment of spacecraft. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 A flowchart of a method for fault detection and isolation of complex equipment on a spacecraft provided in accordance with an embodiment of the present application.

[0017] Figure 2 A schematic diagram of calculating a training loss function provided in one embodiment of the present application.

[0018] Figure 3 A flowchart of a method for fault detection and isolation of complex equipment on a spacecraft is provided as another embodiment of the present application.

[0019] Figure 4 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0021] This application uses time series data of complex spacecraft equipment to perform fault detection and implement fault variable screening. It is a data-driven method that does not require expert knowledge in related fields. In other words, this application implements fault detection and isolation, solving the problem that most methods in the current fault identification field only focus on fault detection. The method of this application can timely determine the variables that cause the fault, which can effectively avoid the risk of fault propagation.

[0022] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0023] In an exemplary embodiment, Figure 1 As shown, a method for fault detection and isolation of complex equipment in a spacecraft is provided. The method is executed by a computer device, and can be specifically executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In an embodiment of the present application, the method includes the following steps 101 to 105.

[0024] Step 101, obtaining current operating status data of complex equipment of a spacecraft; the current operating status data includes a plurality of operating status variables.

[0025] Step 102, input the current operating status data into the operating status prediction model for prediction to obtain the future operating status data distribution; wherein the operating status prediction model is obtained by training the Variational Koopman Anomaly Detector (VKAD) using a time series training sample set of complex equipment of the spacecraft; the time series training sample set includes multiple operating status data within a preset time length; the Variational Koopman Anomaly Detector is determined based on a variational autoencoder and a Koopman matrix layer.

[0026] The variational Koopman anomaly detector includes an input layer, an encoding layer of a variational autoencoder, a Koopman matrix layer, a sampling layer, a decoding layer of a variational autoencoder, and an output layer. The goal of the variational Koopman anomaly detector is to use the encoding layer of the variational autoencoder (corresponding to Figure 2 The q-net network in (a) in the figure encodes the observable distribution of the current running state data, and then propagates the observable distribution forward using an approximate linear operator, and then passes through the decoding layer of the variational autoencoder (corresponding to Figure 2 The p-net network in (a) maps observables back to the state space in the form of probability distribution, thereby achieving uncertainty prediction.

[0027] In the time series training sample set, the working state corresponding to each operating state sample data is that there is no fault, that is, the present application uses the time series data of normal operation (no fault) to train the variational Koopman anomaly detector, so that it can accurately predict the distribution of future data, learn the corresponding mode during normal operation, and then realize the fault detection of the time series data of complex equipment of spacecraft. Obviously, the training of the variational Koopman anomaly detector in the present application is unsupervised training.

[0028] In another embodiment, after training is completed, the operating status prediction model can also be tested using a time series test sample set. If a large deviation occurs after the test sample data is input into the operating status prediction model (because it deviates from the normal mode), it can be considered that a fault has occurred.

[0029] The distribution of future operating state data predicted by the operating state prediction model follows a multivariate Gaussian distribution and is expressed in the form of a predicted probability distribution. For example, the input of the operating state prediction model is a certain time step x(t), and the corresponding probability prediction (u(t+1),σ 2 (t+1)).

[0030] In the process of training the variational Koopman anomaly detector, the training loss function is calculated using the following formula:

[0031]

[0032] Among them, α 1 , α 2 and α 3 is a hyperparameter, indicating the ratio of each corresponding individual loss function to the total loss function; φ is the parameter of the encoding layer of the variational autoencoder, θ is the parameter of the decoding layer of the variational autoencoder, and K is the Kuman operator; is the mapping loss, such as Figure 2 The calculation formula in (b) is as follows:

[0033]

[0034] in, For expectations; is the approximate posterior distribution of the encoding layer output of the variational autoencoder at time t, is the running status data x at time t t The corresponding observables; is the output of the decoding layer of the variational self-encoder and x t The corresponding probability distribution, D KL is the KL divergence, For The corresponding prior distribution, L is the number of running status data in the time series training sample set, To approximate the posterior distribution The lth observable extracted from is the approximate conditional probability distribution; H is the dimension of the function space, which is specified by the designer; and The standard deviation and mean of the hth dimension of the observable distribution at time t, ∈ (l) From the standard Gaussian distribution The lth noise vector extracted from , ⊙ is the element-by-element multiplication.

[0035] is the dynamic loss, which is an additional constraint to ensure that the dynamics of the observables are represented linearly in the function space and controlled by the matrix K. This actually determines a linear Kuman operator K, which is used to promote the evolution of the observable distribution over time, such as Figure 2 The calculation formula of (c) dynamic loss is as follows:

[0036]

[0037] in, is the approximate posterior distribution of the encoding layer output of the variational autoencoder at time t+m, is the running status data x at time t+m t+m The corresponding observable quantity, K m To recursively propagate the m-fold Koopman operator, For The corresponding k-th sampling is, and are the standard deviation and mean of the hth dimension in the observable distribution at time t+m, respectively, where m is the time step; ‖·‖ represents the mean square error, and is averaged over the dimensions.

[0038] To predict the loss, it is possible to make future state predictions, in particular, to represent the predicted future state as a predicted probability distribution, such as Figure 2 As shown in (a) in the figure, the calculation formula for the prediction loss is as follows:

[0039]

[0040] in, For x t+m The corresponding approximate posterior probability distribution is, is a sampling approximation.

[0041] Step 103: Calculate the prediction error matrix and the corresponding T based on the current operation status data and the future operation status data distribution. 2Values, SPE statistics, including:

[0042] The prediction error matrix is ​​calculated according to the following formula: Among them, e t is the prediction error matrix at time t, x t is the operating status data at time t, μ t is the mean value of the observable quantity distribution at time t, where the observable quantity is the predicted future operating status data distribution.

[0043] Calculate T according to the following formula 2 value: in,() T is the transpose; P is the loading matrix, and A is the diagonal matrix consisting of the first k eigenvalues.

[0044] The SPE statistic is calculated according to the following formula: Where I is the identity matrix.

[0045] Step 104: based on the preset limit, according to the T corresponding to the current running status data 2 The working status of the complex equipment of the spacecraft is determined by using the value and the SPE statistic; the working status is a fault or no fault.

[0046] In an application example, the preset limit value includes a first limit value and a second limit value; the first limit value and the second limit value are obtained by training the principal component analysis model based on the prediction error matrix corresponding to each operating status data in the time series training sample set after the training of the variational Koopman anomaly detector is completed.

[0047] The calculation formula of the first limit is:

[0048]

[0049] Among them, T α is the first limit, k is the number of principal components retained in the principal component analysis model, 1-α is the confidence level, and N is the number of samples; F α (k,Nk) follows an F distribution with k and Nk degrees of freedom.

[0050] The calculation formula of the second limit is:

[0051]

[0052] Among them, SPE α is the second limit, m and v are the mean and variance of SPE respectively, Subject to the freedom degree of 2m 2 Chi-square distribution of / v.

[0053] Based on the preset limit, according to the T corresponding to the current operating status data 2 The working state of the complex equipment of the spacecraft is determined by using the value and the SPE statistic, including: when the T corresponding to the current operating state data 2 When the value is greater than the first limit or the SPE statistic is greater than the second limit, the working status of the complex equipment of the spacecraft is marked as a failure.

[0054] Step 105, when the working state is a fault, the deviation score of each operating state variable is calculated according to the current operating state data and the future operating state data distribution, and fault isolation is performed to determine the operating state fault variable.

[0055] The calculation formula for the deviation score of each operating state variable is:

[0056]

[0057] in, is the deviation score of the sth variable in the operating status data at time t, is the sth variable in the running status data at time t, and are the mean and standard deviation of the predicted probability distribution of the sth variable, and |·| is the absolute value. is a positive value, The size of determines the extent to which the variable is affected.

[0058] In an application example, fault isolation is performed to determine operating state fault variables, including: after obtaining the deviation score of each operating state variable, drawing a deviation degree diagram based on the deviation score of each operating state variable to achieve fault isolation; marking the operating state variable with a deviation degree greater than a preset degree value as an operating state fault variable.

[0059] In another exemplary embodiment of the present application, Figure 3As shown, it is divided into two processes: offline modeling and online monitoring. In the offline modeling process, firstly, normal process data is collected and regularized to obtain a time series training sample set; wherein regularization has nothing to do with the distribution of observables, but is only a data preprocessing step for the input model, mainly for maximum-minimization, which is conducive to the training and testing of the model; in this process, the collected time series training sample set is stored in the form of a table with M rows and N columns, the first row and the first column of the table are spaces, the time series training sample set includes N-1 variables, and is stored in the first row of the table except for spaces; the time series training sample set includes M-1 time points, and is stored in the first column of the table except for spaces; the time series training sample set includes (M-1)*(N-1) data points, which are stored in the table except for the first column and the first row; the first row of the mark column of the table is used to store the mark variable, and the other rows of the mark column are used to store the data points collected by the mark variable at different time points; the mark column is any other column of the table except for the first column, and the mark variable is any variable. Then, the variational Koopman anomaly detector is trained to predict the data sample distribution at future time points and calculate the prediction error. After obtaining the prediction error matrix containing all normal samples, the PCA model is trained to determine T 2 and SPE control limits.

[0060] In the online monitoring process, the data samples to be monitored are first collected and regularized, and then the variational Koopman anomaly detector trained in the offline modeling process is used to predict the data samples at the current time point to obtain the data sample distribution at the future time point, that is, Secondly, the prediction error is calculated using the predicted mean and the true sample value, and T 2 The prediction error is monitored by the SPE statistic to see if it exceeds the control limit, thereby determining whether a fault has occurred. If a fault occurs, the predicted mean and standard deviation of the model output are used to set the deviation score to define the degree of deviation of each variable after the fault, and a graph of the degree of deviation of each variable is drawn to achieve fault isolation, thereby determining the fault-related variables.

[0061] In summary, this application studies a fault detection and isolation model based on the Koopman operator theory to address the problems of reliance on the assumptions of linearity and normality of process variables, as well as the difficulty in interpreting the internal processes and decision-making mechanisms of deep learning methods. This model is applicable to complex nonlinear time series data of complex equipment of spacecraft and provides good interpretability. Secondly, in view of the problem that traditional Koopman operator theory does not consider uncertainty, a variational Koopman anomaly detector (VKAD) is studied to perform fault detection and isolation on time series data of complex equipment of uncertain spacecraft. Finally, the mean and standard deviation of the output of the VKAD model help us set up the deviation score, and draw a graph reflecting the degree of deviation of each scalar based on the deviation score, thereby achieving fault isolation.

[0062] The proposed method uses the variational Koopman anomaly detector (VKAD) to detect and isolate faults in time series data of complex spacecraft equipment. VKAD can accurately capture the dynamic characteristics of complex time series and predict the data distribution at future time points based on the data samples at the current time point. After obtaining the predicted distribution of VKAD, the T 2 and SPE monitoring statistics to assess the deviation between the predicted mean and the true sample. If the deviation exceeds T 2 and the control limits of SPE, it is determined to be a faulty sample; after a fault is detected, the predicted mean and standard deviation of the model output are used to set the deviation score to define the degree of deviation of each variable after the fault, and a deviation degree diagram of each variable is drawn to achieve fault isolation, thereby determining the candidate set of fault-related variables.

[0063] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 4 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a fault detection and isolation method for complex equipment of a spacecraft is implemented.

[0064] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0065] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0066] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0067] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0068] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0069] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0070] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0071] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for fault detection and isolation of complex equipment of a spacecraft, characterized in that: The method for fault detection and isolation of complex equipment of a spacecraft comprises: Acquiring current operating state data of complex equipment of a spacecraft; the current operating state data comprising a plurality of operating state variables; The current operating state data is input into the operating state prediction model for prediction to obtain the distribution of future operating state data; wherein the operating state prediction model is obtained by training the variational Koopman anomaly detector using a time series training sample set of complex equipment of the spacecraft; the time series training sample set includes multiple operating state data within a preset time length; the variational Koopman anomaly detector is determined based on a variational autoencoder and a Koopman matrix layer; Based on the current operating state data and the future operating state data distribution, the prediction error matrix and the corresponding T are calculated. 2 Value, SPE statistic; Based on the preset limit, according to the T corresponding to the current operating status data 2 The working state of the complex equipment of the spacecraft is determined by using the value and the SPE statistic; the working state is a fault or no fault; When the working state is a fault, the deviation score of each operating state variable is calculated according to the current operating state data and the future operating state data distribution, and fault isolation is performed to determine the operating state fault variable.

2. The method for fault detection and isolation of complex equipment of a spacecraft according to claim 1, characterized in that: The variational Koopman anomaly detector comprises an input layer, an encoding layer of a variational autoencoder, a Koopman matrix layer, a sampling layer, a decoding layer of a variational autoencoder and an output layer which are arranged in sequence; In the time series training sample set, the working state corresponding to each operating state sample data is that no fault occurs; the distribution of future operating state data predicted by the operating state prediction model follows a multivariate Gaussian distribution and is expressed in the form of a predicted probability distribution.

3. The method for fault detection and isolation of complex equipment of a spacecraft according to claim 2, characterized in that: In the process of training the variational Koopman anomaly detector, the training loss function is calculated using the following formula: Among them, α1, α2 and α3 are hyperparameters; φ is the parameter of the encoding layer of the variational autoencoder, θ is the parameter of the decoding layer of the variational autoencoder, and K is the Kuman operator; For the mapping loss, the calculation formula is as follows: in, For expectations; is the approximate posterior distribution of the encoding layer output of the variational autoencoder at time t, is the running status data x at time t t The corresponding observables; is the output of the decoding layer of the variational self-encoder and x t The corresponding probability distribution, D KL is the KL divergence, For The corresponding prior distribution, L is the number of running status data in the time series training sample set, To approximate the posterior distribution The lth observable extracted from is the approximate conditional probability distribution; H is the dimension of the function space, and The standard deviation and mean of the hth dimension of the observable distribution at time t, ∈ (l) From the standard Gaussian distribution The ,th noise vector extracted from ,⊙ is the bit-by-bit multiplication of the elements; is the dynamic loss, and the calculation formula is as follows: in, is the approximate posterior distribution of the encoding layer output of the variational autoencoder at time t+m, is the running status data x at time t+m t+m The corresponding observable quantity, K m To recursively propagate the m-fold Koopman operator, For The corresponding l-th sampling is and are the standard deviation and mean of the hth dimension in the observable distribution at time t+m, respectively, where m is the time step; ||·|| represents the mean square error, and is averaged over the dimensions; To predict the loss, the calculation formula is as follows: in, For x t+m The corresponding approximate posterior probability distribution is, is a sampling approximation.

4. The method for fault detection and isolation of complex equipment of a spacecraft according to claim 1, characterized in that: The distribution of future operating state data predicted by the operating state prediction model follows a multivariate Gaussian distribution and is expressed in the form of a predicted probability distribution; based on the current operating state data and the future operating state data distribution, the prediction error matrix and the corresponding T 2 Values, SPE statistics, including: The prediction error matrix is ​​calculated according to the following formula: t =|x t -μ t |; Among them, e t is the prediction error matrix at time t, x t is the operating status data at time t, μ t is the mean value of the observable quantity distribution at time t, where the observable quantity is the predicted future operating state data distribution; Calculate T according to the following formula 2 value: in,() T is the transpose; P is the loading matrix, A is the diagonal matrix consisting of the first k eigenvalues; The SPE statistic is calculated according to the following formula:

5. The method for fault detection and isolation of complex equipment of a spacecraft according to claim 1, characterized in that: The preset limit value includes a first limit value and a second limit value; the first limit value and the second limit value are obtained by training the principal component analysis model according to the prediction error matrix corresponding to each operating state data in the time series training sample set after the training of the variational Koopman anomaly detector is completed; Based on the preset limit, according to the T corresponding to the current operating status data 2 The working state of the complex equipment of the spacecraft is determined by using the value and the SPE statistic, including: when the T corresponding to the current operating state data 2 When the value is greater than the first limit or the SPE statistic is greater than the second limit, the working status of the complex equipment of the spacecraft is marked as a failure.

6. The method for fault detection and isolation of complex equipment of a spacecraft according to claim 5, characterized in that: The calculation formula of the first limit is: Among them, T α is the first limit, k is the number of principal components retained in the principal component analysis model, 1-α is the confidence level, and N is the number of samples; F α (k,Nk) follows an F distribution with degrees of freedom k and Nk; The calculation formula of the second limit is: Among them, SPE α is the second limit, m and v are the mean and variance of SPE respectively, Subject to the freedom degree of 2m 2 Chi-square distribution of / v.

7. The method for fault detection and isolation of complex equipment of a spacecraft according to claim 1, characterized in that: The calculation formula for the deviation score of each operating state variable is: in, is the deviation score of the sth variable in the operating status data at time t, is the sth variable in the running status data at time t, and are the mean and standard deviation of the predicted probability distribution of the sth variable, and |·| is the absolute value.

8. The method for fault detection and isolation of complex equipment of a spacecraft according to claim 1, characterized in that: Fault isolation is performed to determine the operating state fault variable, including: after obtaining the deviation score of each operating state variable, drawing a deviation degree diagram according to the deviation score of each operating state variable to achieve fault isolation; marking the operating state variable with a deviation degree greater than a preset degree value as an operating state fault variable.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for fault detection and isolation of complex equipment of a spacecraft as claimed in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for fault detection and isolation of complex equipment of a spacecraft as described in any one of claims 1-8 is implemented.

Images