Lightweight data model adaptive construction method for autonomous diagnosis and reconstruction of spacecraft

Abstract

A lightweight data model adaptive construction method for autonomous diagnosis and reconstruction of a spacecraft comprises the following steps: firstly, constructing a fault mode set and obtaining a telemetry parameter set associated with the fault mode set; secondly, sorting and collecting historical data samples of the telemetry parameters, analyzing the periodicity rule of the telemetry parameters based on Fourier transform, and determining a telemetry parameter subset capable of data modeling; secondly, decomposing the sample according to a periodic term, a trend term and a random term; thirdly, performing statistical analysis on the period items of the historical data samples, and determining data at each moment in a period as a period data template; and finally, performing adaptive fitting on the trend term by adopting an ARIMA method, constructing a trend term characterization model, and superposing the trend term characterization model with the periodic term and the trend term according to moments to form a final data model. Compared with a current mainstream modeling method based on deep learning, the method has the advantages that the data model is lighter, the modeling difficulty is lower, and the method is suitable for data modeling work of spacecrafts represented by massive telemetry parameters.

Description

Technical Field

[0001] The invention relates to a method for adaptively constructing a lightweight data model for autonomous diagnosis and reconstruction of a spacecraft, and belongs to the technical field of overall spacecraft design. Background Art

[0002] How to characterize, determine and quantify the diagnostic and reconfiguration capabilities of the system is a theoretical difficulty and basic problem in the research of autonomous diagnostic and reconfiguration of spacecraft. Spacecraft are composed of multiple subsystems such as control, propulsion, power supply, measurement and control. The system is large in scale, complex in functional structure, and diverse in mission modes. It is difficult to establish its accurate mechanism model under a unified framework. Especially with the development of aerospace technology, spacecraft mission scenarios are more abundant, and more complex components are required to achieve them, such as flexible solar wings, large tanks, and control moment gyroscopes. The coupled vibrations caused by these components make the dynamic characteristics of spacecraft more complex. It is increasingly difficult to accurately characterize the actual system only by relying on the mechanism model, which brings huge challenges to the characterization, determination and quantification of the diagnostic and reconfiguration capabilities of spacecraft. Therefore, it is necessary to make full use of actual engineering data in the process of spacecraft modeling to effectively supplement the mechanism model, improve the accuracy and completeness of the spacecraft system model, and then establish a mathematical expression of diagnosability and reconfigurability under a unified framework to achieve a complete characterization of the diagnostic and reconfiguration capabilities.

[0003] Data modeling based on deep learning methods can achieve high-precision model construction, but on the one hand, the model scale is relatively large and it is difficult to adapt to the needs of on-orbit use. On the other hand, the cost of model modeling is high, requiring a large amount of manpower to adjust hyperparameters and a large amount of computing power to build the model. In addition, during the long-term on-orbit operation of spacecraft, due to the influence of practical factors such as fault propagation, performance degradation, and uncertainty, the data model needs to be continuously updated, and the update work requires manual participation and is difficult to carry out on on-orbit spacecraft with very limited computing resources, which reduces the timeliness of the data model. Therefore, there is an urgent need for a lightweight mathematical model construction method that can adapt to the model, reduce the difficulty of modeling and the scale of the model, and improve the dynamic update capability of the model. Summary of the invention

[0004] The technical problem to be solved by the present invention is: to overcome the deficiencies of the prior art and solve the problem of adaptively constructing a lightweight data model for autonomous diagnosis and reconstruction of spacecraft.

[0005] The purpose of the present invention is achieved through the following technical solutions:

[0006] A lightweight data model adaptive construction method for autonomous diagnosis and reconstruction of spacecraft, comprising:

[0007] (1) Based on the spacecraft system model, the failure cases of each spacecraft system and each single device are studied to construct a complete failure mode set F.

[0008] (2) Establish a one-to-one mapping relationship between each fault mode in the fault mode set obtained in step (1) and the spacecraft telemetry parameters, and then obtain the spacecraft key telemetry parameter set P corresponding to the fault mode set.

[0009] (3) Collect and organize all the key telemetry parameters of the spacecraft in the key telemetry parameter set P obtained in step (2) Historical data sample i is the index number of the fault mode, k is the index number of the telemetry parameter related to a certain fault mode, thereby forming a data sample set D, in which the period of historical data sample collection should be the period when the spacecraft is in normal orbit;

[0010] (4) For the historical data samples formed in (3) Combined with the Fourier transform method, the data in the sample is mapped from the time domain to the frequency domain, and the maximum frequency in the data is counted So as to clarify the telemetry parameter cycle

[0011] (5) According to the telemetry parameter cycle obtained in step (4) The method of centralized moving average is used to convert the trend item data samples From historical data samples Decompose it and use the ARIMA method to automatically complete the data fitting by combining the autocorrelation function (ACF) and partial autocorrelation function (PACF) to form a trend term function x represents the frame sequence of telemetry data.

[0012] (6) From historical data samples Remove the trend item data sample obtained in step (5) And according to the parameter cycle obtained in step (4) The historical data samples without trend data samples are averaged at the same frequency in different periods to form periodic data samples. Then extract the data within one period to form a periodic function

[0013] (7) From historical data samples Remove the trend item data sample obtained in step (5) And the periodic data sample obtained in step (6) Generating random error term data samples And perform mean calculation to form the mean of random error term

[0014] (8) The periodic term function formed from step (5) to step (7) Trend term function Mean of random error term Overlay to form the final data model

[0015] (9) Regularly update the data sample set D constructed in step (3) and repeat steps (4) to (8) to implement the data model Dynamic updates.

[0016] An electronic device, comprising:

[0017] Processor; and

[0018] a memory for storing computer program instructions;

[0019] When the computer program instructions are loaded and executed by the processor, the processor executes the above-mentioned lightweight data model adaptive construction method for autonomous diagnosis and reconstruction of spacecraft.

[0020] A computer-readable storage medium stores computer program instructions, which, when loaded and executed by a processor, enable the processor to execute the above-mentioned method for adaptively constructing a lightweight data model for autonomous diagnosis and reconstruction of a spacecraft.

[0021] A computer program product stored on a non-transitory computer-readable medium includes program codes for implementing the above-mentioned lightweight data model adaptive construction method for autonomous diagnosis and reconstruction of spacecraft.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] (1) Compared with the mainstream deep learning modeling method, this modeling method has a lower modeling cost and does not require human intervention in the modeling process. It can adapt to the automated modeling scenarios of massive telemetry parameters of spacecraft and the model adaptive dynamic update scenarios.

[0024] (2) The data model constructed by this modeling method is composed only of periodic term functions, trend term functions, and the mean of random error terms. The data model is small in scale and is suitable for on-orbit spacecraft with high performance requirements but very limited computing resources.

[0025] (3) The method of the present invention organically combines Fourier transform, data decomposition, ARIMA and other methods, and at the same time ensures the modeling accuracy through targeted modeling of data decomposition and periodic terms, trend terms and random error terms.

[0026] (4) The method of the present invention adopts the ARIMA method to model the trend item data with high modeling cost to ensure the modeling accuracy, and at the same time realizes automatic optimization of the method parameters based on the autocorrelation function (ACF) and the partial autocorrelation function (PACF), thereby improving the modeling efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0028] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0029] A method for adaptively constructing a lightweight data model for autonomous diagnosis and reconstruction of spacecraft, first, construct a set of fault modes and obtain a set of telemetry parameters associated with it; then, organize and collect historical data samples of telemetry parameters, analyze the periodicity of telemetry parameters based on Fourier transform and determine a subset of telemetry parameters that can be used for data modeling; secondly, decompose the samples according to periodic terms, trend terms and random terms; then, perform statistical analysis on the periodic terms of historical data samples, determine the data at each moment in the period, and use it as a periodic data template; finally, use the ARIMA method to adaptively fit the trend terms, construct a trend term representation model, and superimpose it with the periodic terms and trend terms at the moment to form the final data model. Compared with the current mainstream modeling method based on deep learning, the data model of this invention is lighter and less difficult to model, and is suitable for data modeling of spacecraft represented by massive telemetry parameters.

[0030] More specific:

[0031] A lightweight data model adaptive construction method for autonomous diagnosis and reconstruction of spacecraft, such as Figure 1 As shown, including:

[0032] (1) Based on the existing spacecraft failure cases, a complete failure mode set F = {f 0 ,f 1 ,…f n}, n is the total number of failure modes, f i is the i-th failure mode.

[0033] (2) Establish the mapping relationship between each failure mode in the failure mode set obtained in step (1) and the spacecraft telemetry parameter set: i →P i ,i=0,1,2…n,P i is the telemetry parameter set corresponding to the i-th fault mode.

[0034] (3) For each telemetry parameter set All telemetry parameters k represents the kth telemetry parameter involved in the ith fault mode, m i The total number of telemetry parameters involved in the i-th fault mode is collected and sorted to form a data sample set. is the historical data sample corresponding to the kth telemetry parameter involved in the ith fault mode, and x represents the frame sequence in the historical data sample

[0035] (4) Combined with Fourier transform, the telemetry parameters are periodically analyzed and statistically analyzed. Maximum frequency in historical data samples Thus, the telemetry parameter cycle is obtained Right now In each The frame data is one cycle, and the specific formula is as follows:

[0036]

[0037] Among them, argmax(power(freq)) represents the frequency when the maximum amplitude is solved, and power(freq) represents the amplitude when the frequency is freq. for The total amount of data in the data sample.

[0038] (5) According to the telemetry parameter cycle obtained in step (3) The method of centralized moving average is used to convert the trend item data samples from The data samples are decomposed and the ARIMA method is used to automatically complete the data fitting by combining the autocorrelation function (ACF) and the partial autocorrelation function (PACF) to form a trend term function. The specific formula of the centralized moving average is as follows:

[0039]

[0040] (6) From historical data samples Remove the trend item data sample obtained in step (5) Get data sample And according to the parameter cycle obtained in step (4) right Average the data at the same frequency in different periods to form periodic data samples Then extract the data within one period to form a periodic function The specific formula is as follows:

[0041]

[0042] Among them, floor represents the upward rounding function, and ceil represents the upward rounding function.

[0043] (7) From historical data samples Remove the trend item data sample obtained in step (5) And the periodic data sample obtained in step (6) Generating random error term data samples And perform mean calculation to form the mean of random error term The specific formula is as follows:

[0044]

[0045] (8) The periodic term function formed from step (5) to step (7) Trend term function Mean of random error term Overlay to form the final data model

[0046]

[0047] (9) Regularly update the data sample set D constructed in step (3) and repeat steps (4) to (8) to implement the data model Dynamic updates.

[0048] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

[0049] Although the present invention has been disclosed as above in the form of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art may make possible changes and modifications to the technical solution of the present invention by using the methods and technical contents disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope of the technical solution of the present invention.

Claims

1. A lightweight data model adaptive construction method for autonomous diagnosis and reconstruction of spacecraft, characterized in that: include: (1) Based on the spacecraft system model and the failure cases of each system and each single device of the spacecraft, a complete failure mode set F is constructed; (2) establishing a one-to-one mapping relationship between each failure mode in the failure mode set obtained in step (1) and the spacecraft telemetry parameter, thereby obtaining a spacecraft key telemetry parameter set P corresponding to the failure mode set; (3) Collect and organize all the key telemetry parameters of the spacecraft in the key telemetry parameter set P obtained in step (2) Historical data sample i is the index number of the fault mode, k is the index number of the telemetry parameter related to a certain fault mode, thereby forming a data sample set D, in which the period of historical data sample collection should be the period when the spacecraft is in normal orbit; (4) For the historical data samples formed in (3) Using the Fourier transform method, the data in the sample is mapped from the time domain to the frequency domain, and the maximum frequency in the data is counted. So as to clarify the telemetry parameter cycle (5) According to the telemetry parameter cycle obtained in step (4) The trend item data sample From historical data samples Decompose it, and then automatically complete the data fitting to form a trend term function x represents the frame sequence of telemetry data; (6) From historical data samples Remove the trend item data sample obtained in step (5) And data samples And according to the parameter cycle obtained in step (4) The historical data samples without trend data samples are averaged at the same frequency in different periods to form periodic data samples. Then extract the data within one period to form a periodic function (7) From historical data samples Remove the trend item data sample obtained in step (5) And the periodic data sample obtained in step (6) Generating random error term data samples And perform mean calculation to form the mean of random error term (8) The periodic term function formed from step (5) to step (7) Trend term function Mean of random error term Overlay to form the final data model (9) Regularly update the data sample set D constructed in step (3) and repeat steps (4) to (8) to implement the data model Dynamic updates.

2. The method for adaptively constructing a lightweight data model according to claim 1, characterized in that: In step (5), the method of centralized moving average is used to convert the trend item data samples From historical data samples Break it down.

3. The method for adaptively constructing a lightweight data model according to claim 2, characterized in that: The specific formula of the centralized moving average is as follows: Floor represents the rounding function.

4. The method for adaptively constructing a lightweight data model according to claim 1, characterized in that: In step (5), the ARIMA method is used to automatically complete the data fitting by combining the autocorrelation function and the partial autocorrelation function to form the trend term function 5. The method for adaptively constructing a lightweight data model according to claim 1, characterized in that: In step (6): Floor represents the upward rounding function, and ceil represents the upward rounding function.

6. The method for adaptively constructing a lightweight data model according to claim 1, characterized in that: In step (7): for The total amount of data in the data sample.

7. The method for adaptively constructing a lightweight data model according to claim 1, characterized in that: In step (8):

8. An electronic device comprising: processor; as well as a memory for storing computer program instructions; When the computer program instructions are loaded and executed by the processor, the processor performs the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having computer program instructions stored thereon, wherein when the computer program instructions are loaded and executed by a processor, the processor is caused to execute the method according to any one of claims 1 to 7.

10. A computer program product stored on a non-transitory computer readable medium, the computer program product comprising a program code for carrying out the method according to any one of claims 1 to 7.

Images