Telemetering data processing method for spacecraft health assessment

Abstract

The invention relates to a telemetry data processing method for spacecraft health assessment, which comprises the following steps: acquiring in-orbit operation telemetry data of a spacecraft, and performing telemetry decoding according to a data packaging protocol; the decoded data are classified according to the digital quantity and the analog quantity of each subsystem of the spacecraft; processing missing and abnormal data by adopting an ordered nearest neighbor replacement method; carrying out incidence relation analysis on the telemetry data of the spacecraft by utilizing an ID3 algorithm, and carrying out noise filtering and data reconstruction processing by adopting a robust KALMAN filtering method; and analyzing the processed data, calculating the degree of deviation between the processed data and the data in the health state of the part, and evaluating the health state of the spacecraft. According to the data processing method, on-orbit application verification shows that data processing is effective, the health state evaluation detection result conforms to the real situation of on-orbit operation of the spacecraft, and the method has good engineering application value.

Description

Technical Field

[0001] The invention belongs to the technical field of aerospace data processing and relates to a telemetry data processing method for spacecraft health assessment. Background Art

[0002] Telemetry data can accurately reflect the on-orbit performance of a spacecraft. By studying and analyzing the changing patterns of on-orbit spacecraft data and making long-term and short-term predictions of on-orbit telemetry data, we can not only discover the degree of performance degradation of each system component, but also provide early warning and prediction based on data analysis, so that ground-based disposal can be carried out in advance for fault anomalies, reducing the risk of spacecraft operation and ensuring safe and reliable operation of the spacecraft.

[0003] In view of this, how to more effectively mine and utilize on-orbit telemetry data is crucial for the on-orbit maintenance of spacecraft.

[0004] At present, the method for processing spacecraft telemetry data is mainly to decode according to the CCDS international space data standard package format, and perform simple linear fitting, smoothing filtering, and extreme threshold detection processing. It cannot effectively and fully mine and utilize on-orbit telemetry data, nor can it reflect the correlation between data, multi-domain conversion mapping, etc., and it cannot conduct in-depth evaluation and prediction of the health of spacecraft on-orbit operation. Summary of the invention

[0005] The technical problem solved by the present invention is: in view of the fact that the current telemetry data processing method is relatively simple, cannot effectively mine and utilize on-orbit telemetry data, and cannot conduct in-depth assessment and prediction of the health of spacecraft on-orbit operation, an advanced telemetry data processing method for aerospace health assessment is proposed. The method has been verified by application and has good engineering application value.

[0006] The technical solution of the present invention is: a telemetry data processing method for spacecraft health assessment, comprising the following steps:

[0007] Obtain telemetry data of spacecraft in orbit and perform telemetry decoding according to the data packaging protocol;

[0008] Classify the decoded data according to the digital and analog quantities of each subsystem of the spacecraft;

[0009] The missing and abnormal data in each category of classified data are processed by the ordered nearest replacement method;

[0010] The ID3 algorithm is used to analyze the correlation of spacecraft telemetry data processed by the ordered nearest replacement method, and the robust KALMAN filtering method is used to filter out data noise and reconstruct data.

[0011] The data after noise filtering and reconstruction is analyzed, and the degree of deviation from the data under the healthy status of the components is calculated, so as to evaluate the health status of the spacecraft.

[0012] Furthermore, the telemetry decoding is performed according to the data packaging protocol, specifically:

[0013] The spacecraft telemetry data is decoded using an XML-based spacecraft decoding process.

[0014] Furthermore, the classification according to the digital quantities and analog quantities of each subsystem of the spacecraft includes:

[0015] The telemetry data of each subsystem of the spacecraft are classified into digital quantities and analog quantities. Digital quantities include instructions, counts and status, which are further divided into independent digital quantities, associated digital quantities and status digital quantities; analog quantities include current, voltage, angle, temperature and pressure, which are further divided into constant analog quantities, interval analog quantities and trend change analog quantities.

[0016] Furthermore, the processing using the ordered nearest adjacent replacement method includes:

[0017] A state telemetry data sequence is {x t}, t=1,2,3,…n-3,n-2,n-1,n,n+1,n+2,n+3,…, assuming that the telemetry data x n If the data is missing or abnormal, the telemetry data needs to be replaced as soon as possible;

[0018] Yes|x n-3 -x n-2 |、|x n-2 -x n-1 |、|x n+1 -x n+2 |、|x n+2 -x n+3 | sort by size and assign different weights k i (i=n-2,n-1,n+1,n+2), the smaller the value, the greater the weight k i The larger the value, and the k n-2 +k n-1 +k n+1 +k n+2 =1, where |·| represents the difference between two values;

[0019] Then x n =k n-2 *x n-2 +k n-1 *x n-1 +k n+1 *x n+1 +k n+2*x n+2 .

[0020] Furthermore, the correlation analysis includes:

[0021] Let E = F1 × F2 × ... × F n is an n-dimensional finite vector space, where F i , i=1,2,...,n, is a finite discrete symbol set, and the element e in E=(V1,V2,...,V n ) is an example, where V i ∈F i ,i=1,2,...,n;

[0022] Assume that the sizes of the positive example set and the negative example set in the vector space E are p and n respectively, based on the following assumptions:

[0023] ① A correct decision tree on the vector space E has the same classification probability for any sample as the probability of positive and negative examples in E;

[0024] ②The amount of information required for a decision tree to make the correct category judgment for an example is:

[0025]

[0026] If attribute A is the root, it has m values ​​{U1,U2,...,U m}, and divide E into m subsets {E1,E2,...,E m}, assuming E i Contains p i positive examples and n i counterexamples, i = 1, 2, ..., m, subset E i The information entropy is I(p i ,n i ), the information entropy of attribute A is:

[0027]

[0028] The information gain with A as the root is:

[0029] Gain(A)=I(p,n)-E(A)

[0030] Select the attribute A* that maximizes Gain(A) as the extended attribute, and the m subsets E of E corresponding to different values ​​of A* i Recursively call the above process to generate child nodes of A*.

[0031] Furthermore, the use of the robust KALMAN filtering method to perform data noise filtering and data reconstruction processing includes:

[0032] 1) Given the discretized linear equation and observation equation in Kalman filtering:

[0033]

[0034] Among them, x k represents the n×1-order state vector at the kth moment; φ k.k-1 is the n×n order state transfer matrix; w k is the random process vector of n×1 order system; z k represents the m×1-order observation vector; H k is an m×n design matrix; v k is an m×1-order random observation noise vector;

[0035] 2) Constructing the robustness factor model:

[0036]

[0037] Among them, λ k is the weight factor, c is a constant, ranging from 2σ0 to 3σ0, σ0 is the error in the observation noise; s k is the standardized residual:

[0038]

[0039] r k ,σ v They are the prediction residual, redundant observation component and measurement standard deviation of the observed data respectively;

[0040] 3) Constructing adaptive factor model:

[0041]

[0042] a k is the adaptive factor, k is a constant, k = 1.0 ~ 2.5, Δx k Statistics for status mismatch:

[0043]

[0044] in, is the state parameter estimate, is the covariance matrix of the predicted residual vector, tr(·) represents the trace of the matrix;

[0045] 4) After the adaptive factor and robust factor are constructed, the robust adaptive KALMAN filter model is constructed:

[0046]

[0047] The posterior covariance matrix of the state vector is:

[0048]

[0049] in, represents the n×1-order state vector filter estimate at time k; represents the predicted value of the n×1-order state vector at time k; R k is the noise vector v k The covariance matrix of is the predicted value of the m×1-order random observation noise vector; is the transpose of the m×n design matrix; is the covariance matrix of the filter estimation residual vector; is the gain matrix.

[0050] Furthermore, the method for calculating the degree of deviation from the data in the healthy state of the component is as follows:

[0051] According to the design requirements, obtain the maximum and minimum values ​​of the component characteristic quantities in the healthy state of the component:

[0052] Let the i-th characteristic parameter P i The maximum and minimum values ​​are min i and max i ;

[0053] Get the component health status data matrix T under typical health status:

[0054] Assuming that there are K time periods in total, the component health status data matrix T is expressed as an n×K matrix:

[0055]

[0056] The health status data matrix T contains the data at the time when each characteristic parameter takes the extreme value, where: su ,1≤s≤n;1≤u≤K is the characteristic parameter P s Data at time u;

[0057] Standardize the health status data matrix T:

[0058]

[0059] The standardized health status data matrix is ​​denoted as T':

[0060]

[0061] Get the characteristic data of the component health status at a specific moment:

[0062] P obs =(p1,p2,…,p n ) T

[0063] where p i ,i=1,2,…,n represents the characteristic parameter P i Observed value of

[0064] Standardized observation data, standardized feature data is:

[0065]

[0066] The standardized observation data vector is:

[0067] Z obs =(z1,z2,…,z n ) T

[0068] According to the observation vector Z obs and the health state characteristic quantity data matrix T', the degree of component deviation from the health state is calculated as:

[0069] Θ=||[T'·((T') T ·T'+αI K ) -1 (T') T -I n ]·Z obs ||

[0070] In the formula, ||·|| represents the Euclidean norm of the vector, α is the regularization coefficient, and I K is the K×K identity matrix, I n is the n×n identity matrix;

[0071] Calculate the component health status index H:

[0072] The survival function value of the degree of deviation from the healthy state Θ is taken as the quantification of the health degree of the component, and the health degree H is:

[0073]

[0074] The larger the health H value is, the healthier the component is.

[0075] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of a telemetry data processing method for spacecraft health assessment are implemented.

[0076] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of a telemetry data processing method for spacecraft health assessment when executing the computer program.

[0077] A computer program product comprises a computer program, which implements the steps of a telemetry data processing method for spacecraft health assessment when executed by a processor.

[0078] The beneficial effects of the present invention compared with the prior art are:

[0079] (1) The present invention provides a telemetry data processing method for spacecraft health assessment. The method comprehensively considers aspects such as data preprocessing, real-time data correlation analysis, multi-domain data conversion processing, and fault health assessment. The processing method itself is more comprehensive and accurate, and can better tap the application value of the data itself.

[0080] (2) The data processing method proposed in the present invention has been verified in the telemetry data processing of a spacecraft control subsystem. The data processing is effective, and the health status assessment test results are consistent with the actual situation of the spacecraft in orbit, and has good engineering application value.

[0081] (3) The spacecraft telemetry data processing method proposed in the present invention has clear design steps, clear physical meaning, and is easy to implement in engineering. At the same time, the correctness and practicality of the method are verified through example analysis.

[0082] (4) The method proposed in the present invention has obvious functions and simple design. It can be widely applied to the telemetry data processing of on-orbit spacecraft and has strong market competitiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 The figure is a flow chart of an advanced processing method of telemetry data for space health assessment. DETAILED DESCRIPTION

[0084] The present invention will be further described below in conjunction with the embodiments.

[0085] Example 1

[0086] like Figure 1 As shown, this embodiment proposes a telemetry data processing method for spacecraft health assessment, the steps are as follows:

[0087] (1) The spacecraft's on-orbit operation data is transmitted to the ground through the telemetry unit and telemetry decoding is performed according to the data packaging protocol;

[0088] (2) Classify the decoded data according to the digital and analog quantities of each subsystem of the spacecraft;

[0089] (3) After the data are classified, the missing and abnormal data in each category are processed using the ordered nearest replacement method;

[0090] (4) The ID3 algorithm is used to analyze the correlation relationship of the spacecraft telemetry data after the ordered nearest replacement method, and the robust KALMAN filtering method is used to filter out data noise and reconstruct data;

[0091] (5) Analyze the data after noise filtering and reconstruction, calculate the degree of deviation from the data under the healthy status of the components, and thus evaluate the health status of the spacecraft.

[0092] In the step (1), telemetry decoding is performed according to the data packet protocol, which is specifically as follows:

[0093] The XML-based spacecraft decoding and processing method is to use the XML language to define and describe the telemetry data structure of the spacecraft. The characteristic of this method is that it uses a hierarchical and dynamic data definition system to describe the data format, so that the data structure can be dynamically defined like the data, and can be modified according to the system. The processing software can separate the original telemetry data structure by dynamically parsing the XML file structure, and then use these data formats to process the original data. In addition, the language of the XML-based method has a good interface with the database. Most database software supports the parsing of the XML language. Therefore, the application of XML also improves the performance of data storage.

[0094] In this embodiment, the spacecraft telemetry data is decoded using an XML-based spacecraft decoding and processing method, and the data processed by the XML decoding tool presents the status indicators of the on-orbit operation of various systems of the spacecraft.

[0095] In the step (2), the decoded data is classified according to the digital quantity and analog quantity of each subsystem of the spacecraft, specifically:

[0096] Spacecraft usually include thermal control subsystem, attitude and orbit control subsystem, power supply and distribution subsystem, etc. In order to detect the working status of each subsystem, a large number of sensors are installed on the spacecraft to collect various key data and transmit them to the ground. Here, the telemetry data of each subsystem of the spacecraft is classified according to digital quantity and analog quantity.

[0097] Digital quantities can be subdivided into independent digital quantities, associated digital quantities and state digital quantities. Analog quantities can be subdivided into constant analog quantities, interval analog quantities and trend-changing analog quantities. Digital quantities are mainly instructions, counts and states, while analog quantities are mainly current, voltage, angle, temperature, pressure and so on.

[0098] In step (3), the missing and abnormal data in each type of data are processed using the ordered nearest neighbor replacement method, specifically:

[0099] A state telemetry data sequence is {xt} (t=1,2,3,…n-3,n-2,n-1,n,n+1,n+2,n+3,...), assuming that the telemetry data x n If the data is missing or abnormal, the telemetry data needs to be replaced as soon as possible.

[0100] Yes|x n-3 -x n-2 |、|x n-2 -x n-1 |、|x n+1 -x n+2 |、|x n+2 -x n+3 | sort by size and assign different weights k i (i=n-2,n-1,n+1,n+2), the smaller the value, the greater the weight k i The larger the value, and the k n-2 +k n-1 +k n+1 +k n+2 =1;

[0101] where |·| represents the difference between two values.

[0102] Then x n =k n-2 *x n-2 +k n-1 *x n-1 +k n+1 *x n+1 +k n+2 *x n+2 .

[0103] In the step (4), the ID3 algorithm is used to perform correlation analysis on the spacecraft telemetry data processed by the ordered nearest neighbor replacement method, which is specifically as follows:

[0104] In the present invention, the ID3 algorithm is used to perform correlation analysis to mine the knowledge implicit in the telemetry data and the correlation between parameters.

[0105] Let E = F1 × F2 × ... × F n is an n-dimensional finite vector space, where F i , i=1,2,...,n, is a finite discrete symbol set, and the element e in E=(V1,V2,...,V n ) is an example, where V i ∈F i ,i=1,2,...,n。

[0106] Assuming that the sizes of the positive example set and the negative example set in the vector space E are p and n respectively, the ID3 algorithm is based on the following two assumptions:

[0107] ① A correct decision tree on the vector space E has the same classification probability for any sample as the probability of positive and negative examples in E;

[0108] ②The amount of information required for a decision tree to make the correct category judgment for an example is:

[0109]

[0110] If attribute A is the root, it has m values ​​{U1,U2,...,U m}, and divide E into m subsets {E1,E2,…,E m}, here we assume that E i Contains p i positive examples and n i counterexamples, i = 1, 2, ..., m, subset E i The information entropy is I(p i ,n i ), the information entropy of attribute A is:

[0111]

[0112] The information gain with A as the root is:

[0113] Gain(A)=I(p,n)-E(A) (3)

[0114] Select the attribute A* that maximizes Gain(A) as the extended attribute, and the m subsets E of E corresponding to different values ​​of A* i Recursively call the above process to generate child nodes of A*.

[0115] Take the change of the reaction wheel state in the control system as an example to illustrate:

[0116] There are three factors that affect the state change of the reaction wheel: current, voltage, and temperature. After the telemetry data is preprocessed, according to the special requirements of the ID3 algorithm for mining data, the data needs to be discretized. Here, the data change rate is used to discretize the original data. The change rate of the sampling point is recorded as different change intervals to obtain the reaction wheel data training set, as shown in Table 1.

[0117] Table 1 Reaction wheel data training set

[0118] serial number Current (A) Voltage (B) Temperature (C) Reaction wheel abnormality 1 >4％ >6％ >5％ yes 2 >4％ 3％～6％ 2％～5％ yes 3 >4％ <3％ <2％ no 4 4％～2％ >6％ >5％ yes 5 4％～2％ 3％～6％ 2％～5％ no 6 4％～2％ <3％ <2％ no 7 <2％ >6％ >5％ yes 8 <2％ 3％～6％ 2％～5％ no 9 <2％ <3％ <2％ no

[0119] In the training set, the reaction wheel anomaly has two values: yes and no, and therefore corresponds to two different classifications, S1 corresponds to "yes" and S2 corresponds to "no". From Table 1, we know that S1=4 and S2=5.

[0120] The expected information required for classifying a given sample is calculated using formula (1):

[0121]

[0122] Calculate the information entropy of each attribute:

[0123] The current is marked as A:

[0124] For A1>4%, there are 3, S 11 =2,S 21 =1, then:

[0125]

[0126] For A2∈[2%,4%], there are 3, S 12 =1,S 22 =2, then:

[0127]

[0128] For A3<2%, there are 3, S 13 =1,S 23 =2, then:

[0129]

[0130] Therefore, we can calculate from formula (2):

[0131]

[0132] Similarly, voltage is marked as B and temperature is marked as C, and we can calculate E(B)=0.3061, E(C)=0.3061;

[0133] The information gain of the attribute is calculated by formula (3):

[0134] Gain(A)=I(S1,S2)-E(A)=0.0728

[0135] Gain(B)=I(S1,S2)-E(B)=0.6850

[0136] Gain(C)=I(S1,S2)-E(C)=0.6850

[0137] According to the above calculation, voltage and temperature have the same highest information gain and the highest decision tree root node. This shows that voltage and temperature mainly affect the state of the reaction wheel. When monitoring the change of the reaction wheel state, these two telemetry measurements can be focused on.

[0138] In step (4), the robust KALMAN filtering method is used to perform data noise filtering and data reconstruction processing, specifically:

[0139] ①Kalman filtering

[0140] Given the discretized linear equations and observation equations in Kalman filtering:

[0141]

[0142] In the formula, x k represents the n×1-order state vector at the kth moment; φ k.k-1 is the n×n order state transfer matrix; w k is the random process vector of n×1 order system; z k represents the m×1-order observation vector; H k is an m×n design matrix; v k is an m×1 random observation noise vector.

[0143] ②Constructing the robustness factor model

[0144] The determination of the robustness factor is based on the principle of equivalent weights, that is, the robustness factor is constructed through the weight function, and the observations with gross errors are downgraded to reduce the share of abnormal observations in parameter estimation. In order to overcome the blindness of using empirical values ​​in the discrimination conditions and make full use of the current observation information, the discrimination conditions of the Huber weight function are improved:

[0145]

[0146] In the formula, λ k Denoted as the weight factor, c is a constant, generally taken as 2σ0~3σ0 (σ0 is the error in the observation noise).

[0147] and:

[0148]

[0149] In the formula, s k Denoted as standardized residual, r k ,σ v are the prediction residual, redundant observation component and measurement standard deviation of the observed data respectively.

[0150] ③Constructing adaptive factor model

[0151] The adaptive factor model is a return-to-zero function and a non-return-to-zero function. In order to prevent the predicted state covariance from being singular when the adaptive factor is zero, a two-segment function adaptive factor similar to the HUBER function is selected:

[0152]

[0153] In the formula, a k Denoted as adaptive factor, k is a constant, k = 1.0 ~ 2.5, Δx k Statistics for status mismatch:

[0154]

[0155] In the formula, For the state parameter estimation, the least squares solution is generally used. is the covariance matrix of the predicted residual vector, and tr(·) represents the trace of the matrix.

[0156] After the adaptive factor and robust factor are constructed, the robust adaptive KALMAN filter model is constructed:

[0157]

[0158] The posterior covariance matrix of the state vector is:

[0159]

[0160] In the formula, represents the n×1-order state vector filter estimate at time k; represents the predicted value of the n×1-order state vector at time k; R k is the noise vector v k The covariance matrix of is the predicted value of the m×1-order random observation noise vector; is the transpose of the m×n design matrix; is the covariance matrix of the filter estimation residual vector; is the gain matrix.

[0161] After the telemetry data is further processed by the robust KALMAN filtering method, the influence of factors such as sensors, measurement sensors and telemetry signal transmission interference can be effectively removed. The telemetry data is more realistic and can more truly reflect the health characteristics of the spacecraft components themselves.

[0162] In step (5), the degree of deviation from the data in the healthy state of the component is calculated, specifically:

[0163] Assume that the component has n health status characteristic parameters P1, P2, ..., P n ,The steps of component feature extraction and evaluation are as follows:

[0164] ①According to the design requirements, obtain the maximum and minimum values ​​of the component characteristic quantities when the component is in a healthy state.

[0165] Let the i-th characteristic parameter P iThe maximum and minimum values ​​are min i and max i .

[0166] ② Obtain the component health status data matrix T under typical health conditions.

[0167] Assuming that there are K time periods in total, the component health status data matrix T is expressed as an n×K matrix:

[0168]

[0169] The health status data matrix T contains the data at the time when each characteristic parameter takes the extreme value, where: su (1≤s≤n; 1≤u≤K) is the characteristic parameter P s The data at the uth moment.

[0170] The health status data matrix T is standardized using formula (5):

[0171]

[0172] The standardized health status data matrix is ​​denoted as T':

[0173]

[0174] ③ Obtain the characteristic data of the component health status at a specific time:

[0175] P obs =(p1,p2,…,p n ) T (7)

[0176] where p i ,i=1,2,…,n represents the characteristic parameter P i Observed value of .

[0177] ④Use the same method to standardize the observed data, and the standardized feature data is:

[0178]

[0179] The standardized observed data vector is:

[0180] Z obs =(z1,z2,…,z n ) T (9)

[0181] ⑤According to the observation vector Z obs and the health state characteristic quantity data matrix T', the degree of component deviation from the health state is calculated as:

[0182] Θ=||[T'·((T') T ·T'+αI K ) -1 (T') T -I n ]·Z obs || (10)

[0183] In the formula, ||·|| represents the Euclidean norm of the vector, α is the regularization coefficient, and I K is the K×K identity matrix, I n is the n×n identity matrix.

[0184] ⑥ Calculate the component health status index H.

[0185] The survival function value of the degree of deviation from the healthy state Θ is taken as the quantification of the health degree of the component, that is, the health degree H, that is:

[0186]

[0187] The larger the health H value is, the healthier the component is.

[0188] Here the reaction wheel is used as the target for feature evaluation:

[0189] Here, the reaction wheel health characteristic parameters are selected as I, V, T, namely current, voltage and temperature. The minimum and maximum values ​​of the characteristic quantities are:

[0190] I∈[1.1,1.5];V∈[5,10];T∈[20,50];

[0191] Here, the characteristic data of the reaction wheel for a period of time is obtained, and the reaction wheel health status data matrix is ​​obtained according to formula (4):

[0192]

[0193] After the reaction wheel health status data matrix T is standardized using formula (5), we get:

[0194]

[0195] Get the latest observation data:

[0196] P obs =[1.4,7.8,36.2]

[0197] After standardizing the latest observation data obtained using formula (8), we get:

[0198] Z obs =[0.750,0.560,0.540]

[0199] According to formula (10), the degree of component deviation from the healthy state is calculated as:

[0200] θ=2.1425

[0201] Finally, the health of the reaction wheel is calculated according to formula (11):

[0202] H=0.2101

[0203] Therefore, the health status of the reaction wheel can be predicted at all times by calculating the health value of the reaction wheel at all times.

[0204] The present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and when the computer instructions are executed on a computer, the computer executes Figure 1 The method described.

[0205] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) that contain computer-usable program code.

[0206] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0207] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0208] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0209] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

[0210] The contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A telemetry data processing method for spacecraft health assessment, characterized in that: The following steps are involved: Obtain telemetry data of spacecraft in orbit and perform telemetry decoding according to the data packaging protocol; Classify the decoded data according to the digital and analog quantities of each subsystem of the spacecraft; The missing and abnormal data in each category of classified data are processed by the ordered nearest replacement method; The ID3 algorithm is used to analyze the correlation of spacecraft telemetry data processed by the ordered nearest replacement method, and the robust KALMAN filtering method is used to filter out data noise and reconstruct data. The data after noise filtering and reconstruction are analyzed, and the degree of deviation from the data under the healthy status of the components is calculated, so as to evaluate the health status of the spacecraft.

2. A telemetry data processing method for spacecraft health assessment according to claim 1, characterized in that: The telemetry decoding is performed according to the data packaging protocol, specifically: The spacecraft telemetry data is decoded using an XML-based spacecraft decoding process.

3. A telemetry data processing method for spacecraft health assessment according to claim 1, characterized in that: The above classification is based on the digital and analog quantities of each subsystem of the spacecraft, including: The telemetry data of each subsystem of the spacecraft are classified into digital quantities and analog quantities. Digital quantities include instructions, counts and status, which are further divided into independent digital quantities, associated digital quantities and status digital quantities; analog quantities include current, voltage, angle, temperature and pressure, which are further divided into constant analog quantities, interval analog quantities and trend change analog quantities.

4. A telemetry data processing method for spacecraft health assessment according to claim 1, characterized in that: The method of using the ordered nearest replacement method for processing includes: A state telemetry data sequence is {x t }, t=1,2,3,...n-3,n-2,n-1,n,n+1,n+2,n+3,…, assuming that the telemetry data x n If the data is missing or abnormal, the telemetry data needs to be replaced as soon as possible; x n-3 -x n-2 |、|x n-2 -x n-1 |、|x n+1 -x n+2 |、|x n+2 -x n+3 | sort by size and assign different weights k i (i=n-2,n-1,n+1,n+2), the smaller the value, the greater the weight k i The larger the value, and the k n-2 +k n-1 +k n+1 +k n+2 =1, where |·| represents the difference between two values; Then x n =k n-2 *x n-2 +k n-1 *x n-1 +k n+1 *x n+1 +k n+2 *x n+2 .

5. A telemetry data processing method for spacecraft health assessment according to claim 1, characterized in that: The performing of association analysis comprises: Let E = F1 × F2 × ... × F n is an n-dimensional finite vector space, where F i , i=1,2,...,n, is a finite discrete symbol set, and the element e in E=(V1,V2,...,V n ) is an example, where V i ∈F i ,i=1,2,...,n; Assume that the sizes of the positive example set and the negative example set in the vector space E are p and n respectively, based on the following assumptions: ① A correct decision tree on the vector space E has the same classification probability for any sample as the probability of positive and negative examples in E; ②The amount of information required for a decision tree to make the correct category judgment for an example is: If attribute A is the root, it has m values ​​{U1,U2,...,U m }, and divide E into m subsets {E1,E2,...,E m }, assuming E i Contains p i positive examples and n i counterexamples, i = 1, 2, ..., m, subset E i The information entropy is I(p i ,n i ), the information entropy of attribute A is: The information gain with A as the root is: Gain(A)=I(p,n)-E(A) Select the attribute A* that maximizes Gain(A) as the extended attribute, and the m subsets E of E corresponding to different values ​​of A* i Recursively call the above process to generate child nodes of A*.

6. A telemetry data processing method for spacecraft health assessment according to claim 1, characterized in that: The method of using the robust KALMAN filtering method to perform data noise filtering and data reconstruction processing includes: 1) Given the discretized linear equation and observation equation in Kalman filtering: Among them, x k represents the n×1-order state vector at the kth moment; φ k.k-1 is the n×n order state transfer matrix; w k is the random process vector of n×1 order system; z k represents the m×1-order observation vector; H k is an m×n design matrix; v k is an m×1-order random observation noise vector; 2) Constructing the robustness factor model: Among them, λ k is the weight factor, c is a constant, ranging from 2σ0 to 3σ0, σ0 is the error in the observation noise; s k is the standardized residual: r k ,σ v They are the prediction residual, redundant observation component and measurement standard deviation of the observed data respectively; 3) Constructing adaptive factor model: a k is the adaptive factor, k is a constant, k = 1.0 ~ 2.5, Δx k Statistics for status mismatch: in, is the state parameter estimate, is the covariance matrix of the predicted residual vector, tr(·) represents the trace of the matrix; 4) After the adaptive factor and robust factor are constructed, the robust adaptive KALMAN filter model is constructed: The posterior covariance matrix of the state vector is: in, represents the n×1-order state vector filter estimate at time k; represents the predicted value of the n×1-order state vector at time k; R k is the noise vector v k The covariance matrix of is the predicted value of the m×1-order random observation noise vector; is the transpose of the m×n design matrix; is the covariance matrix of the filter estimation residual vector; is the gain matrix.

7. A telemetry data processing method for spacecraft health assessment according to claim 1, characterized in that: The method for calculating the degree of deviation from the data in the healthy state of the component is as follows: According to the design requirements, obtain the maximum and minimum values ​​of the component characteristic quantities in the healthy state of the component: Let the i-th characteristic parameter P i The maximum and minimum values ​​are min i and max i ; Get the component health status data matrix T under typical health status: Assuming that there are K time periods in total, the component health status data matrix T is expressed as an n×K matrix: The health status data matrix T contains the data at the time when each characteristic parameter takes the extreme value, where: y su ,1≤s≤n;1≤u≤K is the characteristic parameter P s Data at time u; Standardize the health status data matrix T: The standardized health status data matrix is ​​denoted as T': Get the characteristic data of the component health status at a specific moment: P obs =(p1,p2,...,p n ) T where p i ,i=1,2,…,n represents the characteristic parameter P i Observed value of Standardized observation data, standardized feature data is: The standardized observation data vector is: WITH obs =(z1,z2,…,z n ) T According to the observation vector Z obs and the health state characteristic quantity data matrix T', the degree of component deviation from the health state is calculated as: Θ=||[T'·((T') T ·T'+αI K ) -1 (T') T -I n ]·Z obs || In the formula, ||·|| represents the Euclidean norm of the vector, α is the regularization coefficient, and I K is the K×K identity matrix, I n is the n×n identity matrix; Calculate the component health status index H: The survival function value of the degree of deviation from the healthy state Θ is taken as the quantification of the health degree of the component, and the health degree H is: The larger the health H value is, the healthier the component is.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 7 are implemented.

Images