CN115563888B - Spacecraft residual life prediction method, system, electronic equipment and medium - Google Patents
Spacecraft residual life prediction method, system, electronic equipment and medium Download PDFInfo
- Publication number
- CN115563888B CN115563888B CN202211552626.4A CN202211552626A CN115563888B CN 115563888 B CN115563888 B CN 115563888B CN 202211552626 A CN202211552626 A CN 202211552626A CN 115563888 B CN115563888 B CN 115563888B
- Authority
- CN
- China
- Prior art keywords
- time
- space
- spacecraft
- data
- residual life
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/04—Ageing analysis or optimisation against ageing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Geometry (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method, a system, electronic equipment and a medium for predicting the residual life of a spacecraft, relating to the technical field of residual life prediction of the spacecraft, wherein the method comprises the following steps: acquiring real-time telemetering time sequence data of the spacecraft; constructing a space-time self-encoder model; the space-time self-encoder model comprises a space-time encoder, a decoder and a residual life prediction module; the space-time encoder comprises a time domain encoder, a space domain encoder and a characteristic fusion module; and inputting the real-time telemetering time sequence data into the trained space-time self-encoder model to obtain the residual life of the spacecraft. The method can improve the prediction precision of the residual life of the spacecraft.
Description
Technical Field
The invention relates to the technical field of spacecraft residual life prediction, in particular to a method, a system, electronic equipment and a medium for predicting the residual life of a spacecraft.
Background
The spacecraft is the most important tool for exploring and developing the universe by human beings at present, and each spacecraft is formed by coupling a large subsystem and a small subsystem, and is very large and complex. In recent decades, with the improvement of science and technology, the aerospace technology has been developed rapidly. More and more countries are aware of the importance of space exploration, with thousands of spacecraft launched into space each year to perform specific scientific tasks. Most spacecrafts are designed at the beginning, a certain 'allowance' is reserved for ensuring the normal operation of the spacecrafts in the space environment, and the spacecrafts launched into the space still can be out of service for a period of time after the spacecrafts are operated for the set time. For example, one fifth of on-orbit spacecrafts in China are in an over-service state, and the number of the over-service spacecrafts is increased along with the lapse of time. How to manage the health of these out-of-service spacecraft is a very important and meaningful thing.
In the health management research of the out-of-service spacecraft, the prediction of the residual service life of the spacecraft is a very important ring. The health state of the in-orbit spacecraft is judged by analyzing the spacecraft telemetering data received by the ground measurement and control center, and possible faults and time of the faults are predicted, so that a certain basis is provided for measurement and control personnel on the ground to timely replace equipment or switch working states. For example, an international space station attitude determination and control office (ACDO) establishes Control Moment Gyro (CMG) early-stage anomaly monitoring software based on an IMS system, and can find an anomaly 14 hours before a CMG fault occurs, so that a worker can replace the CMG with the anomaly to avoid causing greater loss.
The western aerospace strong countries such as the united states research the prediction of the remaining life of the spacecraft as a sub-module of the Prediction and Health Management (PHM), and have achieved good results. In general, the existing remaining life prediction methods can be classified into three major categories, i.e., physical model-based methods, data-driven methods, and hybrid methods. Physical model-based approaches typically require the researcher to have substantial physical and mathematical knowledge and to design the model in conjunction with domain-specific knowledge of the device under study. Such models generally exhibit good results and give clear explanations when the equipment fails and fails. However, the physical model-based method is often used for components or simple systems, and if the system is complex, the design of the model firstly requires huge field knowledge, and even if the system is provided with the condition, the design of the model is an unfinishable task for most engineers due to the complexity of the system. Thus, for complex spacecraft, physical model-based approaches are less feasible.
In recent years, with the development of artificial intelligence technology, a data driving method is more popular, and a good effect on residual life prediction is achieved. The method is characterized in that modeling is carried out according to historical data collected by the sensor, and the degradation process is regarded as a functional relation between the monitoring data of the sensor and the residual service life. The data driving method does not need to rely on domain knowledge for modeling, so that the difficulty of constructing a residual life prediction model of the complex system is reduced.
The mixing method is a method proposed by combining the above two methods, and aims to utilize the advantages of the two methods. However, developing an effective hybrid approach remains very challenging, as it is often not practical for developers to have knowledge both in the domain of the problem and in the data science domain. Therefore, the current mainstream method for predicting the residual life is a data-driven method, and with the development of deep learning technology, the data-driven method for predicting the residual life of the spacecraft is paid more and more attention.
Although the existing deep learning method has a good effect on predicting the residual life of the spacecraft, the problems that the model training process is complex and the interpretability is weak due to the use of a deep network exist, most of deep models only extract the characteristics of one aspect of telemetering time sequence data, in the actual working situation, interaction relations exist among different sensors, the relations change correspondingly with the passage of a working mode or time, and most of the traditional models ignore the acquisition of information of the structural aspect.
Disclosure of Invention
The invention aims to provide a method, a system, electronic equipment and a medium for predicting the residual life of a spacecraft, which can improve the prediction accuracy of the residual life of the spacecraft.
In order to achieve the purpose, the invention provides the following scheme:
a method of spacecraft residual life prediction, the method comprising:
acquiring real-time telemetering time sequence data of the spacecraft;
constructing a space-time self-encoder model; the space-time self-encoder model comprises a space-time encoder, a decoder and a residual life prediction module; the space-time encoder comprises a time domain encoder, a space domain encoder and a characteristic fusion module;
and inputting the real-time telemetering time sequence data into a trained space-time self-encoder model to obtain the residual life of the spacecraft.
Optionally, the training process of the spatio-temporal auto-encoder model includes:
acquiring historical telemetering time sequence data of the spacecraft;
preprocessing the historical telemetering time sequence data to obtain normalized telemetering time sequence data;
manufacturing a remaining life label according to the time sequence of the normalized telemetering time series data;
dividing the normalized telemetering time sequence data with the residual life labels by using a time window method to obtain a plurality of time window data;
and training the space-time self-encoder model by taking the plurality of time window data as input and applying a back propagation algorithm and a minimum loss function to obtain the trained space-time self-encoder model.
Optionally, the preprocessing the historical telemetry time series data to obtain normalized telemetry time series data specifically includes:
filling missing data in the historical telemetering time sequence data by adopting a KNN algorithm to obtain complemented historical telemetering time sequence data;
and carrying out normalization processing on the complemented historical telemetering time sequence data to obtain normalized telemetering time sequence data.
Optionally, the loss function is:
loss=θ×MSE+(1-θ)×RMSE；
wherein the content of the first and second substances,MSEmean square error for remaining service life;RMSEroot mean square error for remaining useful life;θis a weighting factor;x observed is the data value in the remaining life label,x predicted for the predicted residual life of the space-time self-encoder model,Nis the dimension of the time window data.
Optionally, the feature fusion module is a graph neural network.
Optionally, the time domain coding module is a time convolutional network, and is configured to extract time domain information of the input telemetry time series data;
the spatial coding module comprises a graph generator and a graph representation learner; the graph represents a learner as a graph attention network; the graph generator is used for converting the input telemetry time sequence data into a topological graph; the graph representation learner is used for extracting structural information of the topological graph;
the feature fusion module is used for fusing the time domain information and the structural information to obtain fused features;
the decoder is a time convolution network and is used for reducing the size of the fused features into the size of the input telemetering time sequence data to obtain reconstruction data and reconstruction loss;
the residual life prediction module is a long-short term memory network and is used for predicting the residual life according to the fused characteristics to obtain the residual life and the residual life prediction loss;
and determining a loss function according to the reconstruction loss and the predicted loss of the residual life.
A spacecraft residual life prediction system is applied to the spacecraft residual life prediction method, and comprises the following steps:
the acquisition module is used for acquiring real-time telemetering time sequence data of the spacecraft;
the construction module is used for constructing a space-time self-encoder model; the space-time self-encoder model comprises a space-time encoder, a decoder and a residual life prediction module; the space-time encoder comprises a time domain encoder, a space domain encoder and a characteristic fusion module;
and the prediction module is used for inputting the real-time telemetering time sequence data into the trained space-time self-encoder model to obtain the residual life of the spacecraft.
Optionally, the system further comprises a training module; the training module comprises:
the acquisition submodule is used for acquiring historical telemetering time sequence data of the spacecraft;
the normalization submodule is used for preprocessing the historical telemetering time sequence data to obtain normalized telemetering time sequence data;
the marking submodule is used for manufacturing a residual life label according to the time sequence of the normalized telemetering time series data;
the dividing submodule is used for dividing the normalized telemetering time sequence data with the residual life labels by applying a time window method to obtain a plurality of time window data;
and the training submodule is used for training the space-time self-encoder model by taking the plurality of time window data as input and applying a back propagation algorithm and a minimum loss function to obtain the trained space-time self-encoder model.
An electronic device comprises a memory and a processor, wherein the memory is used for storing a computer program, and the processor runs the computer program to enable the electronic device to execute the method for predicting the residual life of the spacecraft.
A computer-readable storage medium, which stores a computer program, which, when executed by a processor, implements the method for predicting the remaining life of a spacecraft as described above.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides a method for predicting the residual life of a spacecraft, which comprises the following steps: acquiring real-time telemetering time sequence data of the spacecraft; constructing a space-time self-encoder model; the space-time self-encoder model comprises a space-time encoder, a decoder and a residual life prediction module; the space-time encoder comprises a time domain encoder, a space domain encoder and a characteristic fusion module; and inputting the real-time telemetering time sequence data into the trained space-time self-encoder model to obtain the residual life of the spacecraft. The invention utilizes the telemetering time sequence data to predict the residual life of the on-orbit spacecraft, and is a data-driven residual life prediction method. The method overcomes the defects that the current method is difficult in feature extraction and only can utilize single-aspect features; and the time domain and space domain (structural aspect) information of the telemetering time sequence data is fully mined by using a space-time self-encoder, and the two kinds of information obtained by mining are input to a prediction module through information fusion, so that the accurate prediction of the residual life of the spacecraft is realized. By the method, ground measurement and control personnel can effectively judge the health condition of the spacecraft in time, so that relevant measures can be taken quickly, the service time of the spacecraft is prolonged, and the loss caused by the early termination of the service life of the spacecraft is reduced.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
Fig. 1 is a flowchart of a method for predicting the remaining life of a spacecraft, provided in an embodiment of the present invention;
FIG. 2 is a schematic diagram of a space-time self-encoder-based spacecraft residual life prediction process provided in an embodiment of the present invention;
FIG. 3 is a schematic diagram of the general architecture of a spatio-temporal self-encoder proposed in the present invention;
fig. 4 is a block diagram of a spacecraft remaining life prediction system provided in an embodiment of the present invention.
Description of the symbols:
1-an obtaining module, 2-a constructing module and 3-a predicting module.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, belong to the protection scope of the present invention.
The invention aims to provide a method, a system, electronic equipment and a medium for predicting the residual life of a spacecraft, which can improve the prediction precision of the residual life of the spacecraft.
The spacecraft residual life prediction method provided by the invention constructs a space-time self-encoder model, the space-time self-encoder model comprises a space-time encoder, a decoder and a residual life prediction module, the space-time encoder acquires time domain and space domain information of data and fuses the time domain and space domain information into new characteristics, the newly fused characteristics comprise time domain information and structural information and have stronger expression capability, and the residual life prediction by using the fused new characteristics has higher accuracy.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Example one
As shown in fig. 1 and fig. 2, the present invention provides a method for predicting the remaining life of a spacecraft, wherein the method comprises:
step S1: and acquiring real-time telemetering time sequence data of the spacecraft.
Step S2: constructing a space-time self-encoder model; the space-time self-encoder model comprises a space-time encoder, a decoder and a residual life prediction module; the space-time encoder comprises a time domain encoder, a space domain encoder and a characteristic fusion module.
And step S3: and inputting the real-time telemetering time sequence data into a trained space-time self-encoder model to obtain the residual life of the spacecraft. Specifically, the real-time telemetering time sequence data is input into a trained space-time self-encoder model to generate reconstruction data and predict the obtained residual life.
In practical application, the spacecraft is arranged at a certain momenttAnd inputting the generated new telemetering data into the space-time self-encoder trained in the step S3, and generating reconstruction data and predicted residual life. Ground measurement and control personnel can use the obtained residual life value as a criterion for the health of the spacecraft.
As a specific implementation manner, the method for predicting the remaining life of the spacecraft, provided by the present invention, further includes a training process of the spatio-temporal auto-encoder model, where the training process specifically includes, after step S2 and before step S3:
step S01: and acquiring historical telemetry time sequence data of the spacecraft.
Step S02: and preprocessing the historical telemetering time sequence data to obtain normalized telemetering time sequence data.
S02 specifically comprises the following steps:
step S021: filling missing data in the historical telemetering time sequence data by adopting a KNN algorithm to obtain complemented historical telemetering time sequence data; in practical application, in the telemetry data acquisition process, due to the fact that technical means and environmental factors tend to cause data to have some deletions, the dimensional value ranges are large in difference and cannot be directly used for model training, residual life labels need to be generated for a data set manually aiming at the missing data aiming at the problem of spacecraft residual life prediction, and a nearest node algorithm (KNN) is used for supplementing the missing data existing in original telemetry data. Identifying similarities or closeness in data sets by calculating distanceskAnd (4) sampling. Then, use thiskSamples to estimate the value of the missing data point. Missing values for each sample are found using the datasetkThe mean of the individual neighborhoods is interpolated.
Step S022: and carrying out normalization processing on the complemented historical telemetering time sequence data to obtain normalized telemetering time sequence data. In particular, the scaling of raw data using linear normalization, whereinX norm In order to be the normalized data, the data,Xas the original data, it is the original data,X max 、X min respectively, the maximum and minimum values of the original data set.
Original data are converted into data in a certain specific range through a maximum value and minimum value principle, so that the influence of data dimension is eliminated. And a linear normalization method is used for carrying out normalization processing on the original telemetering data, so that data of all dimensions are ensured to be in a range, and the training of the model is facilitated.
Step S03: and manufacturing a residual life label according to the time sequence of the normalized telemetering time series data.
In practical application, the remaining life label is made according to the time sequence of the data, and the remaining life label is sequentially decreased to 0 from the initial data to indicate the end of the life. However, the telemetry time series data includes both normal operation data and degradation data. In the normal data stage, all data indexes are always kept stable and basically do not change; only in the degradation stage, the data can show remarkable variation fluctuation, so the invention uses piecewise linear degradation to generate the residual life label, defaults the residual life label to be a fixed value in the stage of data start, and can not start to decline after a certain segmentation point until the life is 0. The determination of the segmentation point is determined according to specific data situations, and the suggested segmentation value is 60% of the total data length.
Step S04: and dividing the normalized telemetering time sequence data with the residual life labels by using a time window method to obtain a plurality of time window data.
Step S05: and training the space-time self-encoder model by taking the plurality of time window data as input and applying a back propagation algorithm and a minimum loss function to obtain the trained space-time self-encoder model. Specifically, the training is completed until the loss function stabilizes.
The method of using the time window divides the normalized telemetering time sequence data with the residual life label into data sections with fixed time step length, thereby facilitating the small batch training of the model and accelerating the training speed of the model. For example, the data is divided by one month for one time window.
FIG. 3 is a schematic diagram of the general architecture of a spatio-temporal self-encoder proposed by the present invention, as shown in FIG. 3, the spatio-temporal self-encoder model specifically includes a spatio-temporal encoder, a decoder and a residual life prediction module; the space-time encoder comprises a time domain encoder, a space domain encoder and a characteristic fusion module.
Wherein, the feature fusion module is a graph neural network.
The time domain coding module is a time convolution network and is used for extracting time domain information of the input telemetering time sequence data.
The spatial coding module comprises a graph generator and a graph representation learner; the graph represents a learner as a graph attention network; the graph generator is used for converting the input telemetry time sequence data into a topological graph; specifically, the number of nodes in the topological graph is determined by the dimensionality of the data, the characteristics of the nodes are the data content of each dimensionality, whether edges exist among the nodes is determined by calculating characteristic correlation among the nodes, if the correlation is larger than a preset threshold value, the edges exist among the nodes, and if not, the edges do not exist; the graph representation learner is configured to extract structural information of the topological graph.
The feature fusion module is used for fusing the time domain information and the structural information to obtain fused features; specifically, the correlation between the features extracted by the time-domain encoder module and the spatial-domain encoding module is calculated, and the calculated correlation is converted into a topological graph structure according to a method similar to that of the spatial-domain encoding module. The neural network of the graph is used for extracting the features, and the fusion of the two features is realized.
The decoder is a time convolution network and is used for reducing the size of the fused features into the size of the input telemetering time sequence data to obtain reconstruction data and reconstruction loss.
And the residual life prediction module is a long-short term memory network and is used for predicting the residual life according to the fused characteristics to obtain the residual life and the residual life prediction loss.
And determining a loss function according to the reconstruction loss and the residual life prediction loss.
The loss function is composed of two parts, one part is reconstruction loss, and the distance is calculated by original data and data reconstructed by a decoder; the other part is the root mean square error of the predicted remaining useful life.
The specific steps of the loss function calculation are as follows:
calculating the dimension of input to a space-time self-encoder asmData of (2)xAnd the dimension generated by the space-time self-encoder ismMean square error between reconstructed data rMSEAnd taking it as a reconstruction error, whereiniIs the first of the dataiMaintaining; in particular, mean square errorMSEThe calculation of (c) is as follows:
calculating root mean square error of remaining useful lifeRMSEIn whichx observed Is the data value in the remaining life label,x predicted predicted remaining life for the model:
and carrying out weighted summation on the obtained MSE and the obtained RMSE, wherein a weight factor theta is used for measuring the proportion of the MSE and the RMSE in the loss function, and the value is between 0 and 1. The loss function of the spatio-temporal autoencoder is as follows:
loss=θ×MSE+(1-θ)×RMSE；
wherein the content of the first and second substances,MSEmean square error for remaining service life;RMSEroot mean square error for remaining service life;θis a weighting factor;x observed is the data value in the remaining life label,x predicted for the predicted remaining life of the spatio-temporal self-encoder model,Nis the dimension of the time window data.
In practical application, processed time window data are simultaneously input into a time domain encoder and a space domain encoder, time domain and space domain information is obtained through processing of the two modules, the obtained time domain and space domain information is synchronous, fusion characteristics are obtained through a GAT-based characteristic fusion module of the time domain and space domain information obtained by the two modules, and the fusion characteristics are simultaneously input into a prediction module formed by a long-short term memory network (LSTM) to obtain the residual life and a decoder to reconstruct the original data. The loss function of the model comprises two parts, wherein one part is residual life prediction loss, the other part is reconstruction loss of data, weight factors are added into the two losses and then the two losses are accumulated, so that the importance degree of the fusion characteristics in a reconstruction task is considered, and the effectiveness of the fusion characteristics in the aspect of residual life prediction is also considered.
The invention provides a spacecraft residual life prediction method, belongs to the cross field of engineering application and information science, and provides a spacecraft residual life prediction method based on a space-time self-encoder from a data driving angle. The method avoids the problem of poor prediction effect caused by overhigh dimensionality of the telemetering data of the spacecraft, and has the advantages of high prediction speed and high accuracy. Ground measurement and control personnel can accurately master the health condition of the spacecraft through the method, and smooth execution of space tasks is ensured. The invention can be used for predicting the residual life of other similar complex systems after being expanded.
Example two
In order to implement the method corresponding to the above embodiment to achieve the corresponding functions and technical effects, the following provides a system for predicting the remaining life of a spacecraft, as shown in fig. 4, the system comprising:
the acquisition module 1 is used for acquiring real-time telemetering time sequence data of the spacecraft.
And the building module 2 is used for building a space-time self-encoder model. The space-time self-encoder model comprises a space-time encoder, a decoder and a residual life prediction module; the space-time encoder comprises a time domain encoder, a space domain encoder and a characteristic fusion module.
And the prediction module 3 is used for inputting the real-time telemetering time sequence data into the trained space-time self-encoder model to obtain the residual life of the spacecraft.
As a specific implementation mode, the spacecraft residual life prediction system further comprises a training module. The training module comprises:
and the acquisition submodule is used for acquiring the historical telemetry time sequence data of the spacecraft.
And the normalization submodule is used for preprocessing the historical telemetering time sequence data to obtain normalized telemetering time sequence data.
And the marking submodule is used for manufacturing the remaining life label according to the time sequence of the normalized telemetering time series data.
And the dividing submodule is used for dividing the normalized telemetering time sequence data with the residual life labels by using a time window method to obtain a plurality of time window data.
And the training submodule is used for training the space-time self-encoder model by taking the plurality of time window data as input and applying a back propagation algorithm and a minimum loss function to obtain the trained space-time self-encoder model.
EXAMPLE III
The embodiment of the invention provides electronic equipment, which comprises a memory and a processor, wherein the memory is used for storing a computer program, and the processor runs the computer program to enable the electronic equipment to execute the method for predicting the residual life of a spacecraft.
Alternatively, the electronic device may be a server.
In addition, an embodiment of the present invention further provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the method for predicting the remaining life of a spacecraft according to the first embodiment of the present invention is implemented.
In the present specification, the embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principle and the embodiment of the present invention are explained by applying specific examples, and the above description of the embodiments is only used to help understanding the method and the core idea of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (8)
1. A method for predicting the residual life of a spacecraft, the method comprising:
acquiring real-time telemetering time sequence data of the spacecraft;
constructing a space-time self-encoder model; the space-time self-encoder model comprises a space-time encoder, a decoder and a residual life prediction module; the space-time encoder comprises a time domain encoder, a space domain encoder and a characteristic fusion module;
inputting the real-time telemetering time sequence data into a trained space-time self-encoder model to obtain the residual life of the spacecraft;
the training process of the space-time self-encoder model comprises the following steps:
acquiring historical telemetering time sequence data of the spacecraft;
preprocessing the historical telemetering time sequence data to obtain normalized telemetering time sequence data;
manufacturing a residual life label according to the time sequence of the normalized telemetering time series data;
dividing the normalized telemetering time sequence data with the residual life labels by using a time window method to obtain a plurality of time window data;
and training the space-time self-encoder model by taking the plurality of time window data as input and applying a back propagation algorithm and a minimum loss function to obtain the trained space-time self-encoder model.
2. The method for predicting the remaining life of a spacecraft according to claim 1, wherein the preprocessing the historical telemetry time series data to obtain normalized telemetry time series data specifically comprises:
filling missing data in the historical telemetering time sequence data by adopting a KNN algorithm to obtain complemented historical telemetering time sequence data;
and carrying out normalization processing on the complemented historical telemetering time sequence data to obtain normalized telemetering time sequence data.
3. A method of predicting the remaining life of a spacecraft as recited in claim 1, wherein said loss function is:
loss=θ×MSE+(1-θ)×RMSE；
wherein, the first and the second end of the pipe are connected with each other,MSEmean square error for remaining service life;RMSEroot mean square error for remaining service life;θis a weight factor;x observed is the data value in the remaining life label,x predicted for the predicted residual life of the space-time self-encoder model,Nis the dimension of the time window data.
4. The method for predicting the remaining life of a spacecraft of claim 1, wherein the feature fusion module is a graph neural network.
5. A spacecraft residual life prediction method according to claim 4, characterized in that the time domain coding module is a time convolution network for extracting time domain information of the input telemetry time series data;
the spatial coding module comprises a graph generator and a graph representation learner; the graph represents a learner as a graph attention network; the graph generator is used for converting the input telemetry time sequence data into a topological graph; the graph representation learner is used for extracting structural information of the topological graph;
the feature fusion module is used for fusing the time domain information and the structural information to obtain fused features;
the decoder is a time convolution network and is used for reducing the size of the fused features into the size of the input telemetering time sequence data to obtain reconstruction data and reconstruction loss;
the residual life prediction module is a long-short term memory network and is used for predicting the residual life according to the fused characteristics to obtain the residual life and the residual life prediction loss;
and determining a loss function according to the reconstruction loss and the residual life prediction loss.
6. A spacecraft remaining life prediction system, the system comprising:
the acquisition module is used for acquiring real-time telemetering time sequence data of the spacecraft;
the construction module is used for constructing a space-time self-encoder model; the space-time self-encoder model comprises a space-time encoder, a decoder and a residual life prediction module; the space-time encoder comprises a time domain encoder, a space domain encoder and a characteristic fusion module;
the prediction module is used for inputting the real-time telemetering time sequence data into a trained space-time self-encoder model to obtain the residual life of the spacecraft;
the system further includes a training module; the training module comprises:
the acquisition submodule is used for acquiring historical telemetering time sequence data of the spacecraft;
the normalization submodule is used for preprocessing the historical telemetering time sequence data to obtain normalized telemetering time sequence data;
the marking submodule is used for manufacturing a residual life label according to the time sequence of the normalized telemetering time series data;
the dividing submodule is used for dividing the normalized telemetering time sequence data with the residual life labels by applying a time window method to obtain a plurality of time window data;
and the training submodule is used for training the space-time self-encoder model by taking the plurality of time window data as input and applying a back propagation algorithm and a minimum loss function to obtain the trained space-time self-encoder model.
7. An electronic device, characterized in that it comprises a memory for storing a computer program and a processor for executing the computer program to make the electronic device execute the method of predicting the remaining life of a spacecraft according to any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that it stores a computer program which, when being executed by a processor, carries out a method for spacecraft residual life prediction according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211552626.4A CN115563888B (en) | 2022-12-06 | 2022-12-06 | Spacecraft residual life prediction method, system, electronic equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211552626.4A CN115563888B (en) | 2022-12-06 | 2022-12-06 | Spacecraft residual life prediction method, system, electronic equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115563888A CN115563888A (en) | 2023-01-03 |
| CN115563888B true CN115563888B (en) | 2023-04-07 |
Family
ID=84770361
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211552626.4A Active CN115563888B (en) | 2022-12-06 | 2022-12-06 | Spacecraft residual life prediction method, system, electronic equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115563888B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117556310B (en) * | 2024-01-10 | 2024-03-26 | 北京中科航天人才服务有限公司 | Spacecraft residual life prediction method |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010188532A (en) * | 2009-02-16 | 2010-09-02 | Ricoh Co Ltd | Image forming apparatus, method of estimating encoder life and program |
| CN102542159A (en) * | 2011-12-08 | 2012-07-04 | 北京空间飞行器总体设计部 | Method for predicting state of on-orbit spacecraft |
| CN102636744A (en) * | 2012-04-24 | 2012-08-15 | 中国科学院空间科学与应用研究中心 | Device and method for detecting relationship between single event effects and time sequence characteristics of FPGA (Field Programmable Gate Array) |
| CN105719037A (en) * | 2014-12-05 | 2016-06-29 | 成都飞机设计研究所 | Aircraft part maintenance and service management method |
| CN109649668A (en) * | 2018-11-12 | 2019-04-19 | 湖北三江航天红峰控制有限公司 | A kind of aircraft health status managing device and method |
| CN114154743A (en) * | 2021-12-13 | 2022-03-08 | 四川大学 | VETMRRN-based space rolling bearing residual life prediction method |
| US11465529B1 (en) * | 2021-06-29 | 2022-10-11 | Beta Air, Llc | Methods and systems for optimizing battery recharge management for use with an electric vertical take-off and landing aircraft |
-
2022
- 2022-12-06 CN CN202211552626.4A patent/CN115563888B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010188532A (en) * | 2009-02-16 | 2010-09-02 | Ricoh Co Ltd | Image forming apparatus, method of estimating encoder life and program |
| CN102542159A (en) * | 2011-12-08 | 2012-07-04 | 北京空间飞行器总体设计部 | Method for predicting state of on-orbit spacecraft |
| CN102636744A (en) * | 2012-04-24 | 2012-08-15 | 中国科学院空间科学与应用研究中心 | Device and method for detecting relationship between single event effects and time sequence characteristics of FPGA (Field Programmable Gate Array) |
| CN105719037A (en) * | 2014-12-05 | 2016-06-29 | 成都飞机设计研究所 | Aircraft part maintenance and service management method |
| CN109649668A (en) * | 2018-11-12 | 2019-04-19 | 湖北三江航天红峰控制有限公司 | A kind of aircraft health status managing device and method |
| US11465529B1 (en) * | 2021-06-29 | 2022-10-11 | Beta Air, Llc | Methods and systems for optimizing battery recharge management for use with an electric vertical take-off and landing aircraft |
| CN114154743A (en) * | 2021-12-13 | 2022-03-08 | 四川大学 | VETMRRN-based space rolling bearing residual life prediction method |
Non-Patent Citations (2)
| Title |
|---|
| 基于DAE-HTPF的新能源汽车电池剩余寿命预测;王正;《机械与电子》;20200324(第03期);5-7+12 * |
| 基于时间序列数据挖掘的航天器故障诊断方法;肇刚等;《飞行器测控学报》;20100615(第03期);5-9 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115563888A (en) | 2023-01-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111784041B (en) | Wind power prediction method and system based on graph convolution neural network | |
| CN105573329A (en) | Attitude and orbit control data analysis method based on decision tree | |
| CN112651437B (en) | Spatial non-cooperative target pose estimation method based on deep learning | |
| CN115563888B (en) | Spacecraft residual life prediction method, system, electronic equipment and medium | |
| CN114049545B (en) | Typhoon intensity determining method, system, equipment and medium based on point cloud voxels | |
| CN116468186B (en) | Flight delay time prediction method, electronic equipment and storage medium | |
| CN113449640A (en) | Remote sensing image building semantic segmentation edge optimization method based on multitask CNN + GCN | |
| CN113779882A (en) | Method, device, equipment and storage medium for predicting residual service life of equipment | |
| CN115690153A (en) | Intelligent agent track prediction method and system | |
| CN115220133A (en) | Multi-meteorological-element rainfall prediction method, device, equipment and storage medium | |
| CN114841072A (en) | Differential fusion Transformer-based time sequence prediction method | |
| CN113627093A (en) | Underwater mechanism cross-scale flow field characteristic prediction method based on improved Unet network | |
| US20230039740A1 (en) | Anomaly detection apparatus, anomaly detection method and program | |
| CN108459933B (en) | Big data computer system fault detection method based on deep recursion network | |
| CN115470957A (en) | System, computer equipment and storage medium for predicting wave height of offshore waves during typhoon based on deep learning | |
| CN112241802A (en) | Interval prediction method for wind power | |
| CN107492129B (en) | Non-convex compressive sensing optimization reconstruction method based on sketch representation and structured clustering | |
| CN116166642A (en) | Spatio-temporal data filling method, system, equipment and medium based on guide information | |
| Medykovskyi et al. | Spectrum neural network filtration technology for improving the forecast accuracy of dynamic processes in economics | |
| Guo et al. | Flight data visualization for simulation & evaluation: a general framework | |
| Zhao et al. | An adaptive online learning model for flight data cluster analysis | |
| CN115618988A (en) | Three-dimensional space-time field joint prediction method for seawater temperature, salinity and flow velocity based on deep neural network | |
| CN113763447A (en) | Method for completing depth map, electronic device and storage medium | |
| CN112862211A (en) | Method and device for assigning orders of dynamic ring defects of communication management system | |
| Zhang | Deep learning-based hybrid short-term solar forecast using sky images and meteorological data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |