Spacecraft health state monitoring method based on recessive characteristics

Abstract

A spacecraft health state monitoring method based on recessive characteristics comprises the following steps: firstly, collecting original telemetering sequences of all telemetering sites of spacecraft key equipment, and preprocessing the original telemetering sequences; then, on the basis of a stacked auto-encoder, coupling features are extracted from telemetry data collected at the same time point, and a recessive degradation feature sequence is obtained; secondly, smooth filtering is carried out on the recessive degradation feature sequence, and a Gaussian curve fitting sequence and an SG curve fitting sequence are obtained respectively; and finally, respectively calculating root-mean-square errors between the recessive degradation feature sequence and Gaussian and SG curve fitting sequences, and selecting the curve fitting sequence with the small root-mean-square error as a health state modeling result. According to the method, coupling characteristics of all telemetering site data are extracted through a stacked auto-encoder, and a model capable of truly reflecting the health state of the spacecraft is obtained by using a filter and a residual evaluation index, so that the problem that the health state of the spacecraft is difficult to accurately monitor in a complex environment and in different degradation stages is solved.

Description

Technical Field

[0001] This invention relates to a method for monitoring the health status of spacecraft based on latent features, belonging to the field of aerospace. Background Technology

[0002] As highly reliable systems, spacecraft require careful study of the performance evolution patterns and processes of their critical equipment and monitoring their health status. This is crucial for ensuring the safe, reliable, and stable operation of spacecraft in orbit. However, due to the harsh conditions of outer space, the performance degradation of critical equipment is influenced by multiple factors. Existing methods typically rely on historical knowledge and engineering experience, pre-selecting a fitting function to fit a curve of the spacecraft's health status. This approach is challenging, as the evolution process cannot be uniformly described by a single fitting function, accurately representing the impact of multiple factors on critical equipment, and failing to account for the differences in different degradation stages. Furthermore, existing methods can only utilize data from a single telemetry point for health status curve fitting, failing to integrate complex data from multiple telemetry points to represent the health status of critical equipment. Therefore, existing methods cannot accurately reflect the true health status of critical equipment. Summary of the Invention

[0003] The technical problem solved by this invention is to overcome the shortcomings of existing technologies and provide a spacecraft health status monitoring method based on implicit features. This method extracts the coupling features of all telemetry points through stacked autoencoders, and simultaneously uses filters and residual evaluation indicators to adaptively obtain a model that can truly reflect the spacecraft's health status, thus solving the problem that existing methods cannot accurately reflect the true health status.

[0004] The technical solution of this invention is: A method for monitoring the health status of spacecraft based on latent features, comprising the following steps: (1) Collect key equipment of spacecraft m The original telemetry sequences of each telemetry site are generated from the original telemetry sequences of the corresponding telemetry site at time points 1, 2, ..., 3. n The collected telemetry data are arranged from earliest to latest according to the collection time; the original telemetry sequences of all telemetry sites are preprocessed, and the telemetry dataset is constructed using the preprocessed telemetry data of all telemetry sites. (2) Based on the stacked autoencoder, for the first telemetry point, the second telemetry point, ..., the ... m Telemetry data collected from multiple telemetry sites at the same time point are correlated and coupled, and coupling features are extracted. Then, implicit degradation feature sequences of key spacecraft equipment are extracted from the telemetry dataset. ;in, HIi For the first telemetry point, second telemetry point, ..., the output of the stacked autoencoder m At time point, each telemetry point i The results of coupled feature extraction from the collected telemetry data, i.e., the key equipment of the spacecraft at a given time point... i Latent degradation characteristics; (3) The implicit degradation feature sequence is smoothed and filtered to obtain the Gaussian implicit degradation feature curve fitting sequence and the SG implicit degradation feature curve fitting sequence of the key equipment of the spacecraft respectively; the Gaussian implicit degradation feature curve fitting sequence is formed by arranging and combining the implicit degradation features of the key equipment of the spacecraft at all time points in the order of Gaussian filtering and in the order of time points from early to late; the SG implicit degradation feature curve fitting sequence is formed by arranging and combining the implicit degradation features of the key equipment of the spacecraft at all time points in the order of time points from early to late after SG filtering. (4) Calculate the root mean square error between the latent degradation feature sequence and the Gaussian latent degradation feature curve fitting sequence. Calculate the root mean square error between the latent degradation feature sequence and the fitted sequence of the SG latent degradation feature curve. ;like Less than If the condition is met, the Gaussian latent degradation characteristic curve fitting sequence is used as the spacecraft health status change curve, and the spacecraft health status is monitored through the health status change curve; otherwise, the SG latent degradation characteristic curve fitting sequence is used as the spacecraft health status change curve, and the spacecraft health status is monitored through the health status change curve.

[0005] Furthermore, step (1), which involves preprocessing the original telemetry sequences of all telemetry sites, is as follows: (1.1) Based on the Lagrange interpolation method, the missing values ​​in the original telemetry sequences of all telemetry sites are interpolated and compensated to obtain a consistent telemetry sequence of all telemetry sites. (1.2) Using the Wright criterion, outliers are removed from the consistent telemetry sequences of each telemetry site to obtain the final telemetry sequence of each telemetry site. The telemetry sequences of all telemetry sites constitute the telemetry dataset.

[0006] Furthermore, the specific process of interpolating and compensating for missing values ​​in the telemetry sequences of all telemetry sites in step (1.1) is as follows: (1.1.1) Interpolate and compensate for missing values ​​in the telemetry sequence of the first telemetry site, using the position index of each element in the sequence as the x-axis. The element's own value is used as the y-axis. ; (1.1.2) Using the x-coordinates of the known numerical elements in the telemetry sequence, the elements with known numerical values ​​are those with non-missing values; construct a Lagrange basis function for each element with known numerical values, the expression of which is:

[0007] in, For the first k Lagrange basis functions for n known numerical elements, , These are the first telemetry sequences. k -1、 k The x-coordinates of +1 known numerical elements, and the total number of known numerical elements in the telemetry sequence are: L indivual; (1.1.3) The ordinates of the known numerical elements in the telemetry sequence With the corresponding basis functions Weighted summation yields the interpolation polynomial. P ( x The expression is

[0008] (1.1.4) Substitute the x-coordinates of the missing values ​​in the telemetry sequence into the interpolation polynomial. P ( x In the process, the interpolation compensation result of the corresponding missing value element is calculated; the interpolation compensation result is used to replace the missing value element in the original telemetry sequence to obtain the consistent telemetry sequence of the first telemetry site. (1.1.5) Sequentially acquire the second telemetry point, the third telemetry point, ..., the ... m The consistent telemetry sequence of each telemetry site is obtained by following the same steps (1.1.1) to (1.1.4).

[0009] Furthermore, the specific steps for outlier removal in step (1.2) of the consistent telemetry sequence for each telemetry site are as follows: (1.2.1) Perform sliding window processing on the consistent telemetry sequence of the first telemetry site; the sliding window processing is performed with a window size of... a The sliding window moves along the sequence of consistent telemetry data, each... a A series of consecutive values ​​form a sliding window sequence; if the number of elements at the end of the consistent telemetry sequence is insufficient... a Each telemetry point is grouped into a separate group, with the number of elements matching the number of end elements; the consistent telemetry sequence of the first telemetry point is divided into [number] groups. b A sliding window sequence, and all sliding window sequences constitute the sliding window sequence set for the corresponding telemetry site; (1.2.2) Calculate the mean of the first sliding window sequence of the first telemetry site. and standard deviation ; (1.2.3) Calculate the residuals of all elements in the first sliding window sequence of the first telemetry point using the following formula:

[0010] in, For the first sliding window in the sequence i The residual of each element, For the first i One element; (1.2.4) Ratio the residual of each element in the first sliding window sequence to the left and right. Compare; if the residual of an element is greater than If the index of a given element is consecutive, it is considered a pending outlier; otherwise, it is considered a normal value. If three or more pending outliers have consecutive indices in the sliding window sequence, these three or more pending outliers with consecutive indices are considered normal values; otherwise, the pending outlier is considered an outlier and removed from the consistent telemetry sequence of the first telemetry point. The time when the first telemetry point collected the outlier is recorded and defined as the outlier time. The second telemetry point, ..., ... m Data collected at each telemetry station during out-of-field time were synchronously removed; (1.2.5) For the second sliding window sequence, ..., the third sliding window sequence in the sliding window sequence set of the first telemetry site b The outlier removal is performed on each sliding window sequence, following the same procedure as steps (1.2.2) to (1.2.4). (1.2.6) For the second telemetry site, the third telemetry site, ..., the ... m Outlier removal is performed on the consistent telemetry sequences of each telemetry site, following the same procedure as steps (1.2.1) to (1.2.5).

[0011] Furthermore, the specific process of extracting the latent degradation feature sequence of key spacecraft equipment from the telemetry dataset in step (2) is as follows: (2.1) Input the telemetry sequences of all telemetry points into the stacked autoencoder; the stacked autoencoder has three encoders and three decoders; (2.2) The stacked autoencoder performs multi-layer encoding on the telemetry sequences of all telemetry points, encoding the first telemetry point, the second telemetry point, ..., the ... m Telemetry data collected from several telemetry sites at the same time point are correlated and coupled, and coupling features are extracted. The final output consists of two data points of length [length missing]. nThe initial latent degradation feature sequence of key spacecraft equipment; the initial latent degradation feature sequence with a more trend-like nature was selected as the latent degradation feature sequence of key spacecraft equipment. ;in, HI i For the first telemetry point, second telemetry point, ..., the output of the stacked autoencoder m At time point, each telemetry point i The results of coupling feature extraction from the acquired telemetry sequences, i.e., the key equipment of the spacecraft at specific time points... i The latent degenerative characteristics.

[0012] Furthermore, the specific steps for smoothing the latent degenerate feature sequence in step (3) are as follows: (3.1) Gaussian filtering is applied to the latent degenerate feature sequence, and the calculation formula is as follows:

[0013] in, HI i The first recessive degenerative feature sequence i The latent degradation characteristics of key spacecraft equipment at specific points in time. For the first i The result after Gaussian filtering of the latent degradation characteristics of key spacecraft equipment at a given time point. The Gaussian distribution parameters of the Gaussian filter are used; all the implicit degradation features after Gaussian filtering are arranged in order from earliest to latest time to form a Gaussian implicit degradation feature curve fitting sequence. (3.2) Perform SG filtering on the latent degenerate feature sequence. The calculation formula is as follows:

[0014] in, The window width set for SG filtering, For smoothing coefficients, The first recessive degenerative feature sequence i+k The latent degradation characteristics of key spacecraft equipment at specific points in time. For the first i The result of SG filtering on the latent degradation characteristics of key spacecraft equipment at a given time point.

[0015] Furthermore, in step (4), the root mean square error of the fitted sequence of the latent degradation feature sequence and the Gaussian latent degradation feature curve is... The calculation formula is:

[0016] in, and These are the latent degradation feature sequences and the Gaussian latent degradation feature curve fitting sequences, respectively. i One element; The root mean square error of the fitted sequence of the latent degradation feature sequence and the SG latent degradation feature curve in step (4) The calculation formula is:

[0017] in, and These are the latent degradation feature sequences and the SG latent degradation feature curve fitting sequences, respectively. i Each element.

[0018] Secondly, the present invention also proposes a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the aforementioned spacecraft health status monitoring method based on latent features.

[0019] Thirdly, the present invention also proposes a processor, characterized in that the processor is used to run a program, wherein the program executes the above-described method for monitoring the health status of a spacecraft based on latent features.

[0020] Fourthly, a storage medium comprising a stored program, wherein, during the execution of the program, the device containing the storage medium is controlled to execute the aforementioned spacecraft health status monitoring method based on latent features.

[0021] The advantages of this invention compared to the prior art are: (1) This invention extracts the latent performance degradation features of key spacecraft equipment by stacking autoencoders, thereby coordinating and coupling the data of all telemetry points, and can accurately and comprehensively describe the performance degradation of key spacecraft equipment.

[0022] (2) This invention combines filters and residual evaluation indicators to fit the health status curve of key equipment, and adaptively obtains a model that can truly reflect the health status of spacecraft, so as to solve the problem that it is difficult to accurately monitor the health status of spacecraft under complex environment and different degradation stages, and can more accurately represent the health fluctuation of key equipment of spacecraft. Attached Figure Description

[0023] Figure 1 This is a flowchart of a spacecraft health status monitoring method based on latent features according to the present invention; Figure 2 This is a diagram of the stacked autoencoder network structure used in the spacecraft health status monitoring method based on latent features of the present invention. Detailed Implementation

[0024] The specific embodiments of the present invention will now be described in further detail with reference to the accompanying drawings.

[0025] like Figure 1 As shown, this invention provides a spacecraft health status monitoring method based on latent features, the steps of which are as follows: (1) Collect key equipment of spacecraft m The original telemetry sequences of each telemetry site are generated from the original telemetry sequences of the corresponding telemetry site at time points 1, 2, ..., 3. n The collected telemetry data are arranged from earliest to latest according to the collection time; the original telemetry sequences of all telemetry sites are preprocessed, and the telemetry dataset is constructed using the preprocessed telemetry data of all telemetry sites. The step (1) of preprocessing the original telemetry sequences of all telemetry sites is as follows: (1.1) Based on the Lagrange interpolation method, missing values ​​in the original telemetry sequences of all telemetry sites are interpolated and compensated to obtain a consistent telemetry sequence for all telemetry sites. The specific process is as follows: (1.1.1) Interpolate and compensate for missing values ​​in the telemetry sequence of the first telemetry site, using the position index of each element in the sequence as the x-axis. The element's own value is used as the y-axis. ; (1.1.2) Using the x-coordinates of the known numerical elements in the telemetry sequence, the elements with known numerical values ​​are those with non-missing values; construct a Lagrange basis function for each element with known numerical values, the expression of which is:

[0026] in, For the first k Lagrange basis functions for n known numerical elements, , These are the first telemetry sequences. k -1、 k The x-coordinates of +1 known numerical elements, and the total number of known numerical elements in the telemetry sequence are: L indivual; (1.1.3) The ordinates of the known numerical elements in the telemetry sequence With the corresponding basis functions Weighted summation yields the interpolation polynomial. P ( x The expression is

[0027] (1.1.4) Substitute the x-coordinates of the missing values ​​in the telemetry sequence into the interpolation polynomial. P ( x In the process, the interpolation compensation result of the corresponding missing value element is calculated; the interpolation compensation result is used to replace the missing value element in the original telemetry sequence to obtain the consistent telemetry sequence of the first telemetry site. (1.1.5) Sequentially acquire the second telemetry point, the third telemetry point, ..., the ... m The consistent telemetry sequence of each telemetry site is obtained by following the same steps (1.1.1) to (1.1.4).

[0028] (1.2) Using the Wright criterion, outlier removal is performed on the consistent telemetry sequences of each telemetry site to obtain the final telemetry sequence for each telemetry site. The telemetry sequences of all telemetry sites constitute the telemetry dataset. The specific steps are as follows: (1.2.1) Perform sliding window processing on the consistent telemetry sequence of the first telemetry site; the sliding window processing is performed with a window size of... a The sliding window moves along the sequence of consistent telemetry data, each... a A series of consecutive values ​​form a sliding window sequence; if the number of elements at the end of the consistent telemetry sequence is insufficient... a Each telemetry point is grouped into a separate group, with the number of elements matching the number of end elements; the consistent telemetry sequence of the first telemetry point is divided into [number] groups. b A sliding window sequence, and all sliding window sequences constitute the sliding window sequence set for the corresponding telemetry site; (1.2.2) Calculate the mean of the first sliding window sequence of the first telemetry site. and standard deviation ; (1.2.3) Calculate the residuals of all elements in the first sliding window sequence of the first telemetry point using the following formula:

[0029] in, For the first sliding window in the sequence i The residual of each element, For the first i One element; (1.2.4) Ratio the residual of each element in the first sliding window sequence to the left and right. Compare; if the residual of an element is greater than If the index of a given element is consecutive, it is considered a pending outlier; otherwise, it is considered a normal value. If three or more pending outliers have consecutive indices in the sliding window sequence, these three or more pending outliers with consecutive indices are considered normal values; otherwise, the pending outlier is considered an outlier and removed from the consistent telemetry sequence of the first telemetry point. The time when the first telemetry point collected the outlier is recorded and defined as the outlier time. The second telemetry point, ..., ... m Data collected at each telemetry station during out-of-field time were synchronously removed; (1.2.5) For the second sliding window sequence, ..., the third sliding window sequence in the sliding window sequence set of the first telemetry site b The outlier removal is performed on each sliding window sequence, following the same procedure as steps (1.2.2) to (1.2.4). (1.2.6) For the second telemetry site, the third telemetry site, ..., the ... m Outlier removal is performed on the consistent telemetry sequences of each telemetry site, following the same procedure as steps (1.2.1) to (1.2.5).

[0030] (2) Based on the stacked autoencoder, for the first telemetry point, the second telemetry point, ..., the ... m Telemetry data collected from multiple telemetry sites at the same time point are correlated and coupled, and coupling features are extracted. Then, implicit degradation feature sequences of key spacecraft equipment are extracted from the telemetry dataset. ;in, HI i For the first telemetry point, second telemetry point, ..., the output of the stacked autoencoder m At time point, each telemetry point i The results of coupled feature extraction from the collected telemetry data, i.e., the key equipment of the spacecraft at a given time point... i Latent degradation characteristics; The specific process of extracting the latent degradation feature sequence of key spacecraft equipment from the telemetry dataset in step (2) is as follows: (2.1) Input the telemetry sequences of all telemetry points into the stacked autoencoder; the stacked autoencoder has three encoders and three decoders, and the network structure is as follows: Figure 2 As shown in Table 1, the network-related configurations are as follows: The dimensions of the input and output data; Table 1 Stacked Autoencoder Network Configuration

[0031] (2.2) The stacked autoencoder performs multi-layer encoding on the telemetry sequences of all telemetry points, encoding the first telemetry point, the second telemetry point, ..., the ... mTelemetry data collected from several telemetry sites at the same time point are correlated and coupled, and coupling features are extracted. The final output consists of two data points of length [length missing]. n The initial latent degradation feature sequence of key spacecraft equipment; the initial latent degradation feature sequence with a more trend-like nature was selected as the latent degradation feature sequence of key spacecraft equipment. ;in, HI i For the first telemetry point, second telemetry point, ..., the output of the stacked autoencoder m At time point, each telemetry point i The results of coupling feature extraction from the acquired telemetry sequences, i.e., the key equipment of the spacecraft at specific time points... i Latent degradation characteristics; Through the above process, this invention extracts the latent performance degradation characteristics of key spacecraft equipment based on stacked autoencoders, thereby correlating and coupling the data of all telemetry points, and can accurately and comprehensively describe the performance degradation of key spacecraft equipment.

[0032] (3) The implicit degradation feature sequence is smoothed and filtered to obtain the Gaussian implicit degradation feature curve fitting sequence and the SG implicit degradation feature curve fitting sequence of the key equipment of the spacecraft respectively; the Gaussian implicit degradation feature curve fitting sequence is formed by arranging and combining the implicit degradation features of the key equipment of the spacecraft at all time points in the order of Gaussian filtering and in the order of time points from early to late; the SG implicit degradation feature curve fitting sequence is formed by arranging and combining the implicit degradation features of the key equipment of the spacecraft at all time points in the order of time points from early to late after SG filtering. The specific steps for smoothing the latent degenerate feature sequence in step (3) are as follows: (3.1) Gaussian filtering is applied to the latent degenerate feature sequence, and the calculation formula is as follows:

[0033] in, HI i The first recessive degenerative feature sequence i The latent degradation characteristics of key spacecraft equipment at specific points in time. For the first i The result after Gaussian filtering of the latent degradation characteristics of key spacecraft equipment at a given time point. The Gaussian distribution parameters of the Gaussian filter are used; all the implicit degradation features after Gaussian filtering are arranged in order from earliest to latest time to form a Gaussian implicit degradation feature curve fitting sequence. (3.2) Perform SG filtering on the latent degenerate feature sequence. The calculation formula is as follows:

[0034] in, The window width set for SG filtering, For smoothing coefficients, The first recessive degenerative feature sequence i+k The latent degradation characteristics of key spacecraft equipment at specific points in time. For the first i The result of SG filtering on the latent degradation characteristics of key spacecraft equipment at a given time point.

[0035] (4) Calculate the root mean square error between the latent degradation feature sequence and the Gaussian latent degradation feature curve fitting sequence. Calculate the root mean square error between the latent degradation feature sequence and the fitted sequence of the SG latent degradation feature curve. ;like Less than If the condition is met, the Gaussian latent degradation characteristic curve fitting sequence is used as the spacecraft health status change curve, and the spacecraft health status is monitored through the health status change curve; otherwise, the SG latent degradation characteristic curve fitting sequence is used as the spacecraft health status change curve, and the spacecraft health status is monitored through the health status change curve.

[0036] The root mean square error of the fitted sequence of the latent degradation feature sequence and the Gaussian latent degradation feature curve in step (4) The calculation formula is:

[0037] in, and These are the latent degradation feature sequences and the Gaussian latent degradation feature curve fitting sequences, respectively. i One element; The root mean square error of the fitted sequence of the latent degradation feature sequence and the SG latent degradation feature curve in step (4) The calculation formula is:

[0038] in, and These are the latent degradation feature sequences and the SG latent degradation feature curve fitting sequences, respectively. i Each element.

[0039] Through the above process, this invention combines filters and residual evaluation indicators to fit the health status curves of key equipment, and adaptively obtains a model that can truly reflect the health status of spacecraft. This addresses the challenge of accurately monitoring the health status of spacecraft under complex environments and different degradation stages, and can more accurately represent the health fluctuations of key spacecraft equipment.

[0040] Secondly, the present invention also proposes a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the aforementioned spacecraft health status monitoring method based on latent features.

[0041] Thirdly, the present invention also proposes a processor, characterized in that the processor is used to run a program, wherein the program executes the above-described method for monitoring the health status of a spacecraft based on latent features.

[0042] Fourthly, a storage medium comprising a stored program, wherein, during the execution of the program, the device containing the storage medium is controlled to execute the aforementioned spacecraft health status monitoring method based on latent features.

[0043] The parts of this invention not described in detail are common knowledge to those skilled in the art.

Claims

1. A method for monitoring the health status of spacecraft based on latent features, characterized in that... The steps are as follows: (1) Collect key equipment of spacecraft m The original telemetry sequences of each telemetry site are generated from the original telemetry sequences of the corresponding telemetry site at time points 1, 2, ..., 3. n The collected telemetry data are arranged from earliest to latest according to the collection time; the original telemetry sequences of all telemetry sites are preprocessed, and the telemetry dataset is constructed using the preprocessed telemetry data of all telemetry sites. (2) Based on the stacked autoencoder, for the first telemetry point, the second telemetry point, ..., the ... m Telemetry data collected from multiple telemetry sites at the same time point are correlated and coupled, and coupling features are extracted. Then, implicit degradation feature sequences of key spacecraft equipment are extracted from the telemetry dataset. ;in, HI i For the first telemetry point, second telemetry point, ..., the output of the stacked autoencoder m At time point, each telemetry point i The results of coupled feature extraction from the collected telemetry data, i.e., the key equipment of the spacecraft at a given time point... i Latent degradation characteristics; (3) The implicit degradation feature sequence is smoothed and filtered to obtain the Gaussian implicit degradation feature curve fitting sequence and the SG implicit degradation feature curve fitting sequence of the key equipment of the spacecraft respectively; the Gaussian implicit degradation feature curve fitting sequence is formed by arranging and combining the implicit degradation features of the key equipment of the spacecraft at all time points in the order of Gaussian filtering and in the order of time points from early to late; the SG implicit degradation feature curve fitting sequence is formed by arranging and combining the implicit degradation features of the key equipment of the spacecraft at all time points in the order of time points from early to late after SG filtering. (4) Calculate the root mean square error between the latent degradation feature sequence and the Gaussian latent degradation feature curve fitting sequence. Calculate the root mean square error between the latent degradation feature sequence and the fitted sequence of the SG latent degradation feature curve. ;like Less than If the condition is met, the Gaussian latent degradation characteristic curve fitting sequence is used as the spacecraft health status change curve, and the spacecraft health status is monitored through the health status change curve; otherwise, the SG latent degradation characteristic curve fitting sequence is used as the spacecraft health status change curve, and the spacecraft health status is monitored through the health status change curve.

2. The spacecraft health status monitoring method based on latent features according to claim 1, characterized in that: The step (1) of preprocessing the original telemetry sequences of all telemetry sites is as follows: (1.1) Based on the Lagrange interpolation method, the missing values ​​in the original telemetry sequences of all telemetry sites are interpolated and compensated to obtain a consistent telemetry sequence of all telemetry sites. (1.2) Using the Wright criterion, outliers are removed from the consistent telemetry sequences of each telemetry site to obtain the final telemetry sequence of each telemetry site. The telemetry sequences of all telemetry sites constitute the telemetry dataset.

3. The spacecraft health status monitoring method based on latent features according to claim 2, characterized in that: The specific process of interpolating and compensating for missing values ​​in the telemetry sequences of all telemetry sites in step (1.1) is as follows: (1.1.1) Interpolate and compensate for missing values ​​in the telemetry sequence of the first telemetry site, using the position index of each element in the sequence as the x-axis. The element's own value is used as the y-axis. ; (1.1.2) Using the x-coordinates of the known numerical elements in the telemetry sequence, the elements with known numerical values ​​are those with non-missing values; construct a Lagrange basis function for each element with known numerical values, the expression of which is: in, For the first k Lagrange basis functions for n known numerical elements, , These are the first telemetry sequences. k -1、 k The x-coordinates of +1 known numerical elements, and the total number of known numerical elements in the telemetry sequence are: L indivual; (1.1.3) The ordinates of the known numerical elements in the telemetry sequence With the corresponding basis functions Weighted summation yields the interpolation polynomial. P ( x The expression is (1.1.4) Substitute the x-coordinates of the missing values ​​in the telemetry sequence into the interpolation polynomial. P ( x In the process, the interpolation compensation result of the corresponding missing value element is calculated; the interpolation compensation result is used to replace the missing value element in the original telemetry sequence to obtain the consistent telemetry sequence of the first telemetry site. (1.1.5) Sequentially acquire the second telemetry point, the third telemetry point, ..., the ... m The consistent telemetry sequence of each telemetry site is obtained by following the same steps (1.1.1) to (1.1.4).

4. The spacecraft health status monitoring method based on latent features according to claim 2, characterized in that: The specific steps for outlier removal in step (1.2) of the consistent telemetry sequence for each telemetry site are as follows: (1.2.1) Perform sliding window processing on the consistent telemetry sequence of the first telemetry site; the sliding window processing is performed with a window size of... a The sliding window moves along the sequence of consistent telemetry data, each... a A series of consecutive values ​​form a sliding window sequence; if the number of elements at the end of the consistent telemetry sequence is insufficient... a Each telemetry point is grouped into a separate group, with the number of elements matching the number of end elements; the consistent telemetry sequence of the first telemetry point is divided into [number] groups. b A sliding window sequence, and all sliding window sequences constitute the sliding window sequence set for the corresponding telemetry site; (1.2.2) Calculate the mean of the first sliding window sequence of the first telemetry site. and standard deviation ; (1.2.3) Calculate the residuals of all elements in the first sliding window sequence of the first telemetry point using the following formula: in, For the first sliding window in the sequence i The residual of each element, For the first i One element; (1.2.4) Ratio the residual of each element in the first sliding window sequence to the left and right. Compare; if the residual of an element is greater than If the index of a given element is consecutive, it is considered a pending outlier; otherwise, it is considered a normal value. If three or more pending outliers have consecutive indices in the sliding window sequence, these three or more pending outliers with consecutive indices are considered normal values; otherwise, the pending outlier is considered an outlier and removed from the consistent telemetry sequence of the first telemetry point. The time when the first telemetry point collected the outlier is recorded and defined as the outlier time. The second telemetry point, ..., ... m Data collected at each telemetry station during out-of-field time were synchronously removed; (1.2.5) For the second sliding window sequence, ..., the third sliding window sequence in the sliding window sequence set of the first telemetry site b The outlier removal is performed on each sliding window sequence, following the same procedure as steps (1.2.2) to (1.2.4). (1.2.6) For the second telemetry site, the third telemetry site, ..., the ... m Outlier removal is performed on the consistent telemetry sequences of each telemetry site, following the same procedure as steps (1.2.1) to (1.2.5).

5. The spacecraft health status monitoring method based on latent features according to claim 1, characterized in that: The specific process of extracting the latent degradation feature sequence of key spacecraft equipment from the telemetry dataset in step (2) is as follows: (2.1) Input the telemetry sequences of all telemetry points into the stacked autoencoder; the stacked autoencoder has three encoders and three decoders; (2.2) The stacked autoencoder performs multi-layer encoding on the telemetry sequences of all telemetry points, encoding the first telemetry point, the second telemetry point, ..., the ... m Telemetry data collected from several telemetry sites at the same time point are correlated and coupled, and coupling features are extracted. The final output consists of two data points of length [length missing]. n The initial latent degradation feature sequence of key spacecraft equipment; the initial latent degradation feature sequence with a more trend-like nature was selected as the latent degradation feature sequence of key spacecraft equipment. ;in, HI i For the first telemetry point, second telemetry point, ..., the output of the stacked autoencoder m At time point, each telemetry point i The results of coupling feature extraction from the acquired telemetry sequences, i.e., the key equipment of the spacecraft at specific time points... i The latent degenerative characteristics.

6. The spacecraft health status monitoring method based on latent features according to claim 1, characterized in that: The specific steps for smoothing the latent degenerate feature sequence in step (3) are as follows: (3.1) Gaussian filtering is applied to the latent degenerate feature sequence, and the calculation formula is as follows: in, HI i The first recessive degenerative feature sequence i The latent degradation characteristics of key spacecraft equipment at specific points in time. For the first i The result after Gaussian filtering of the latent degradation characteristics of key spacecraft equipment at a given time point. The Gaussian distribution parameters of the Gaussian filter are used; all the implicit degradation features after Gaussian filtering are arranged in order from earliest to latest time to form a Gaussian implicit degradation feature curve fitting sequence. (3.2) Perform SG filtering on the latent degenerate feature sequence. The calculation formula is as follows: in, The window width set for SG filtering, For smoothing coefficients, The first recessive degenerative feature sequence i+k The latent degradation characteristics of key spacecraft equipment at specific points in time. For the first i The result of SG filtering on the latent degradation characteristics of key spacecraft equipment at a given time point.

7. The spacecraft health status monitoring method based on latent features according to claim 1, characterized in that: The root mean square error of the fitted sequence of the latent degradation feature sequence and the Gaussian latent degradation feature curve in step (4) The calculation formula is: in, and These are the latent degradation feature sequences and the Gaussian latent degradation feature curve fitting sequences, respectively. i One element; The root mean square error of the fitted sequence of the latent degradation feature sequence and the SG latent degradation feature curve in step (4) The calculation formula is: in, and These are the latent degradation feature sequences and the SG latent degradation feature curve fitting sequences, respectively. i Each element.

8. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the spacecraft health status monitoring method based on latent features as described in any one of claims 1 to 7.

9. A processor, characterized in that, The processor is used to run a program, wherein the program executes a spacecraft health status monitoring method based on latent features as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium includes a stored program, wherein, when the program is executed, the device containing the storage medium is controlled to perform a spacecraft health status monitoring method based on latent features as described in any one of claims 1 to 7.

Images