Spacecraft micro-vibration source separation method and fault prediction method

Abstract

The invention discloses a spacecraft micro-vibration source separation method and a fault prediction method, and the method comprises the steps: carrying out the feature extraction of micro-vibration data in the in-orbit operation process of a spacecraft, obtaining a feature map, endowing each channel of the feature map with a weight value, obtaining a feature vector, mapping the feature vector into probability distribution, and carrying out the fault prediction of the micro-vibration source. And separation and category matching of micro-vibration sources are realized. Based on the micro-vibration source obtained through separation, faults of a spacecraft single machine and a platform can be further predicted in combination with historical data or telemetering abnormal conditions are proved and analyzed, so that the accuracy and efficiency of spacecraft fault prediction are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of spacecraft fault diagnosis, and in particular to a method for separating a spacecraft micro-vibration source and a method for predicting a fault. Background Art

[0002] As human space activities become increasingly frequent and complex, the requirements for the stability and safety of spacecraft on-orbit operations are becoming increasingly stringent. Spacecraft on-orbit face a variety of external interferences, such as atmospheric disturbances, changes in Earth's gravity, and solar radiation pressure. These interferences can easily cause varying degrees of micro-vibration in the spacecraft. Furthermore, micro-vibrations generated by the operation of individual components within the spacecraft, such as SADA solar orientation, the QV antenna power-up and rotation, and the thermal control cooling unit startup, can also affect the overall micro-vibration of the spacecraft. On the one hand, these micro-vibrations can interfere with the spacecraft's attitude control and orbit maintenance, and may also damage the spacecraft's internal structures and equipment, leading to failures. On the other hand, micro-vibrations also provide a key observation window for insight into the spacecraft's current operating status, helping to verify the working conditions of the spacecraft and its internal components and predict the probability of potential failures.

[0003] Traditional methods for spacecraft fault detection and diagnosis are primarily categorized as knowledge-driven and data-driven. Knowledge-driven methods, such as expert systems, detect faults based on explicit rules. Currently, the mainstream knowledge-driven approach to spacecraft fault detection relies on knowledge and experience derived from telemetry measurements of various subsystems, particularly analyzing telemetry data from the power supply, propulsion, and integrated electronics subsystems. Data-driven methods, on the other hand, utilize various machine learning techniques to analyze faults from data and are particularly suitable for detecting unknown faults.

[0004] To ensure the normal operation of spacecraft, real-time monitoring and analysis of its micro-vibration status is extremely critical. Traditional micro-vibration analysis mainly relies on ground physical simulation models and signal processing technology. A large amount of data is collected during the ground single-machine and spacecraft mechanical environment test phases, and the micro-vibration situation is analyzed with the help of simulation models. However, this method is mainly used to eliminate manufacturing and early process defects in harsh mechanical environments, or to verify fatigue and wear life. Signal processing technology is centered on Fourier transform, which realizes signal decomposition and noise reduction through time domain and frequency domain conversion, and plays a certain role in micro-vibration signal analysis. However, in the face of complex and changeable micro-vibration signals, existing methods have significant limitations. First, in terms of fault detection and diagnosis, existing technologies mostly use knowledge-driven methods and rely heavily on large amounts of satellite-to-ground telemetry data. When key telemetry data is insufficiently transmitted or abnormal, it is difficult to accurately judge the status of the spacecraft. In addition, knowledge-driven methods overly rely on expert experience. In today's giant communication constellations containing tens of thousands of satellites and increasingly rich space activities, it is difficult to achieve scientific management and control solely by manpower; and in micro-vibration signal processing, although existing technologies can extract some features, they cannot accurately identify and separate different micro-vibration interference sources, and it is difficult to effectively classify complex micro-vibration sources affected by atmospheric disturbances, space debris, and the power-on and power-off interference of single-unit equipment within the device. Summary of the Invention

[0005] In order to address some or all of the problems in the prior art, the present invention provides, in a first aspect, a method for separating a spacecraft microvibration source, comprising:

[0006] Extract features from the micro-vibration data of the spacecraft during on-orbit operation and obtain a feature map;

[0007] Assigning a weight value to each channel of the feature map to obtain a feature vector; and

[0008] The feature vector is mapped into a probability distribution to achieve separation and category matching of micro-vibration sources.

[0009] Furthermore, the micro-vibration data is pre-processed before feature extraction.

[0010] Furthermore, the preprocessing includes denoising and data format standardization.

[0011] Furthermore, feature extraction is performed through filters composed of multiple convolutional layers.

[0012] Furthermore, a weight value is assigned to each channel of the feature map based on an attention mechanism.

[0013] Furthermore, the feature vector is mapped to a probability distribution through the Softmax layer.

[0014] Based on the separation method described above, a second aspect of the present invention provides a spacecraft fault prediction method based on deep learning, comprising:

[0015] Collect micro-vibration data of spacecraft during on-orbit operation;

[0016] Separating the micro-vibration data by the aforementioned separation method to obtain micro-vibration source data; and

[0017] The fault state of the spacecraft is predicted based on the micro-vibration source data through a fault prediction model, wherein the fault prediction model is deployed on the spacecraft.

[0018] Furthermore, the fault prediction model is constructed through the following steps:

[0019] Evaluate the impact of different micro-vibration sources on spacecraft performance based on micro-vibration source data; and

[0020] Through deep learning technology, a fault prediction model is built based on the analysis of evaluation results and historical fault data.

[0021] Furthermore, the spacecraft fault prediction method further includes:

[0022] The fault prediction model is trained and optimized based on historical micro-vibration data.

[0023] Furthermore, the spacecraft fault prediction method further includes:

[0024] sending the predicted fault status to a ground station; and

[0025] The ground station analyzes the operating status of the spacecraft based on the data from the onboard temperature sensor and pressure sensor and the fault status.

[0026] A third aspect of the present invention provides an onboard autonomous health management system, comprising:

[0027] A data acquisition module, which is used to collect micro-vibration data of the spacecraft during its on-orbit operation;

[0028] a micro-vibration source separation module, which is communicatively connected to the data acquisition module and is used to separate the micro-vibration source based on the micro-vibration data to obtain micro-vibration source data; and

[0029] A fault prediction module is communicatively connected to the micro-vibration source separation module and is used to predict vibration rationality based on the micro-vibration source data and predict the fault state of the spacecraft.

[0030] The present invention provides a method for separating spacecraft microvibration sources and a fault prediction method, which achieves precise analysis of spacecraft microvibrations on-orbit through a special microvibration interference source separation algorithm. The separation method can effectively separate different microvibration sources and microvibration intensities, thereby accurately determining the impact of microvibrations on the spacecraft. At the same time, the fault prediction method can predict faults of individual spacecraft and platforms based on historical data or perform corroborative analysis of telemetry anomalies, providing reliable data support capabilities. The fault prediction method can improve the accuracy and efficiency of spacecraft fault prediction and provide strong technical support for the safe operation of spacecraft. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] To further illustrate the above and other advantages and features of various embodiments of the present invention, a more detailed description of various embodiments of the present invention will be presented with reference to the accompanying drawings. It will be understood that these drawings depict only typical embodiments of the present invention and are not to be considered as limiting the scope thereof. In the drawings, for clarity, identical or corresponding components will be represented by the same or similar reference numerals.

[0032] Figure 1 A schematic diagram showing the vibration sources of a spacecraft during its on-orbit operation;

[0033] Figure 2 A schematic flow chart showing a method for separating a spacecraft micro-vibration source according to an embodiment of the present invention;

[0034] Figure 3 A schematic structural diagram of a spacecraft micro-vibration source separation model according to an embodiment of the present invention is shown;

[0035] Figure 4 A schematic diagram illustrating a flow chart of a method for predicting spacecraft faults based on deep learning according to an embodiment of the present invention; and

[0036] Figure 5 A schematic structural diagram of an onboard autonomous health management system according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0037] In the following description, the present invention is described with reference to various embodiments. However, those skilled in the art will recognize that the embodiments may be implemented without one or more of the specific details or with other alternative and / or additional methods or components. In other cases, well-known structures, materials, or operations are not shown or described in detail to avoid obscuring the inventive aspects of the present invention. Similarly, specific numbers and configurations are set forth for illustrative purposes in order to provide a comprehensive understanding of the embodiments of the present invention. However, the present invention is not limited to these specific details.

[0038] In this specification, reference to "one embodiment" or "the embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. The appearances of the phrase "in one embodiment" in various places in this specification are not necessarily all referring to the same embodiment.

[0039] It should be noted that the embodiments of the present invention describe the method steps in a specific order, but this is only for the purpose of illustrating the specific embodiment and does not limit the order of the steps. On the contrary, in different embodiments of the present invention, the order of the steps can be adjusted according to actual needs.

[0040] In the present invention, each module of the system according to the present invention can be implemented using software, hardware, firmware or a combination thereof. When the module is implemented using software, the function of the module can be implemented by a computer program flow, for example, the module can be implemented by a code segment (such as a code segment in a language such as C, C++) stored in a storage device (such as a hard disk, a memory, etc.), wherein when the code segment is executed by a processor, the corresponding function of the module can be implemented. When the module is implemented using hardware, the function of the module can be implemented by setting a corresponding hardware structure, for example, by hardware programming a programmable device such as a field programmable gate array (FPGA) to implement the function of the module, or by designing an application-specific integrated circuit (ASIC) including electronic devices such as a plurality of transistors, resistors and capacitors to implement the function of the module. When the module is implemented using firmware, the function of the module can be written into a read-only memory such as an EPROM or EEPROM of the device in the form of program code, and when the program code is executed by the processor, the corresponding function of the module can be implemented. In addition, certain functions of the module may need to be implemented by separate hardware or through collaboration with the hardware, for example, the detection function is implemented by corresponding sensors (such as proximity sensors, accelerometers, gyroscopes, etc.), the signal transmission function is implemented by corresponding communication devices (such as Bluetooth devices, infrared communication devices, baseband communication devices, Wi-Fi communication devices, etc.), the output function is implemented by corresponding output devices (such as displays, speakers, etc.), and so on.

[0041] like Figure 1 As shown in the figure, during the on-orbit operation of a spacecraft, it will be affected by various types of micro-vibrations. The micro-vibrations include special vibrations generated when there is a mechanical failure in the spacecraft, which can be recorded as fault vibration U e , superimposed on the micro-vibration of the spacecraft in normal motion to form the micro-vibration U v , the fault vibration collected by various precision collection sensors of the spacecraft, such as the vibration U collected by the momentum wheel t , the vibration U collected by the star sensor s and the vibration U collected by the optical fast mirror in the laser payloadk After research, the inventor found that the collected vibration U v 、U t 、U s and U k Restore to fault vibration U e , and can classify various fault vibrations through deep learning methods.

[0042] The essence of fault vibration classification through deep learning methods lies in the learning of vibration formula by deep learning algorithms. The vibration formula can be derived through the following process. If the posture is represented by Euler quaternion [q1 q2 q3 q4] T If it is stipulated that:

[0043]

[0044] F Aits =R sAi F Ait ,

[0045] Among them, R Ais For R sAi For F Ait For m Aik For Aik For F Aits for

[0046] Then, the dynamic equation of the flexible body micro-vibration relative to the spacecraft itself is:

[0047]

[0048] By performing data science-based feature learning on the spacecraft's sensor data, we essentially find the best fitting model through gradient descent of high-dimensional data features.

[0049] Based on the above principles, the present invention provides a deep learning-based spacecraft fault prediction method and system. This method uses a unique microvibration interference source separation algorithm to accurately analyze on-orbit microvibrations of spacecraft, thereby providing a basis for fault prediction. The technical solution of the present invention is further described below with reference to the accompanying drawings.

[0050] Figure 2 A schematic flow chart showing a method for separating a spacecraft micro-vibration source according to an embodiment of the present invention is shown. Figure 2 As shown, a method for separating a spacecraft micro-vibration source includes:

[0051] First, in step 201, data preprocessing is performed. The micro-vibration data collected by various sensors during the on-orbit operation of the spacecraft are preprocessed. Since there are many micro-vibration sources on the spacecraft, different micro-vibrations are usually collected by different sensors, such as micro-vibration sensors and other precision collection sensors, among which precision collection sensors include momentum wheels, star sensors, and optical fast-reflecting mirrors in laser payloads. The data formats output by different sensors may be different. In addition, there may be certain noise in the data collection process. Therefore, in order to improve the accuracy and reliability of the data, it is necessary to first preprocess the collected micro-vibration data, such as denoising, standardization and other operations, to eliminate noise and unify the data format.

[0052] Next, in step 202, feature extraction is performed on the micro-vibration data of the spacecraft during its on-orbit operation to obtain a feature map. Figure 3 FIG. 1 is a schematic diagram showing the structure of a spacecraft micro-vibration source separation model according to an embodiment of the present invention. Figure 3 As shown, in one embodiment of the present invention, a deep neural network structure is constructed by multi-layer convolution and attention mechanism to achieve separation of micro-vibration sources, wherein the multi-layer convolution layer extracts features from the pre-processed micro-vibration data through filters. Figure 3 As shown in the figure, after the data is input, it is first converted into a continuous vector representation through input embedding, and then enters the multi-layer convolution layer, which includes several alternating 3*3 convolution layers (convolutions) and maximum pooling layers (Max-Pool), as well as several Inception modules and an average pooling layer (Avg-Pool). The convolution layer performs a convolution operation on the input data by sliding the convolution kernel, which can automatically extract local features in the data. With the increase of convolution layers, more advanced and abstract features can be extracted. The maximum pooling layer is used to select the maximum value in each pooling area as the most important feature, retaining the main information of the data. It reduces the dimension of the data by downsampling the input data, reduces the computational complexity and storage requirements of the model, helps to prevent overfitting and improve the generalization ability of the model. Average pooling can smooth the features and reduce the impact of noise. The Inception module extracts features of different scales by using multiple convolution kernels and pooling operations of different sizes in parallel.

[0053] Next, in step 203, weights are assigned. A weight value is assigned to each channel of the feature map to obtain a feature vector. As mentioned above, in one embodiment of the present invention, a deep neural network structure is constructed by multi-layer convolution and attention mechanism to achieve separation of micro-vibration sources, wherein the attention mechanism (Attention) is used to screen the importance of each channel of the acquired feature map, and assign a corresponding weight value to each feature based on the importance, so that the network can focus on feature channels that are useful for the task, thereby improving the channels of the feature map that are useful for the current task, and suppressing feature channels that are not very useful for the current task; and

[0054] Finally, in step 204, the micro-vibration sources are separated. The feature vectors are mapped into probability distributions to achieve separation and category matching of micro-vibration sources. In one embodiment of the present invention, Figure 3 As shown in the figure, the features extracted by the previous layers are first integrated through the fully connected layer (FC) to obtain a higher-level and more abstract feature representation. Then, the feature vector is mapped to a function of another vector through the Softmax layer. The element value of the output vector represents the probability distribution of each category, thereby achieving the separation of different micro-vibration sources and category matching.

[0055] In one embodiment of the present invention, the accuracy and generalization ability of the separation method can be improved by optimizing a large amount of training data.

[0056] Based on the separated micro-vibration source data, the impact of different micro-vibration sources on spacecraft performance can be further evaluated, providing data support for subsequent fault prediction and elimination.

[0057] Figure 4 FIG1 is a flow chart showing a method for predicting spacecraft faults based on deep learning according to an embodiment of the present invention. Figure 4 As shown, a spacecraft fault prediction method based on deep learning includes:

[0058] First, in step 401, micro-vibration data is collected. The micro-vibration data collected during the spacecraft's on-orbit operation includes collecting vibration U through the momentum wheel. t , star sensor collects vibration U s , the optical part of the laser payload fast mirror collects vibration U k , and the micro-vibration U collected by the micro-vibration sensor v Micro-vibration data of spacecraft during on-orbit operation;

[0059] Next, in step 402, the micro-vibration source is separated. The micro-vibration data is separated by the separation method described above to obtain micro-vibration source data; and

[0060] Finally, in step 403, the fault state is predicted. A fault prediction model is used to predict the spacecraft's fault state based on the microvibration source data, where the fault prediction model is deployed on the spacecraft. As previously described, the impact of the spacecraft's microvibration can be assessed based on the isolated microvibration source data. By comparing and analyzing the impact of different microvibration sources on spacecraft performance and analyzing and learning from historical fault data, the spacecraft's failure trends can be predicted based on the microvibration source data.

[0061] In one embodiment of the present invention, the fault prediction model is constructed using deep learning technology and based on the results of a microvibration impact assessment. Specifically, after processing the microvibration data source, the impact of microvibration on the spacecraft is evaluated. By comparing and analyzing the impact of different microvibration sources on spacecraft performance, data support can be provided for subsequent fault prediction and troubleshooting. Based on the microvibration impact assessment results, a fault prediction model can be constructed. This fault prediction model uses deep learning technology to predict spacecraft failure trends through analysis and learning from historical fault data. By continuously updating and optimizing model parameters, the accuracy and efficiency of fault prediction can also be improved.

[0062] In one embodiment of the present invention, to improve the accuracy and generalization capability of the fault prediction model, the model can be trained using historical microvibration data, and its parameters can be optimized through cross-validation. The trained fault prediction model is uploaded to the spacecraft's onboard computer via a satellite-to-ground link.

[0063] The separation methods and fault prediction models mentioned above are all based on neural networks or deep learning technologies. Therefore, they can continuously adapt to changes in the spacecraft operating environment and the emergence of new micro-vibration interference sources, and maintain the ability of autonomous learning and self-optimization of vibrations and faults.

[0064] Based on the separation method and fault prediction method described above, Figure 5 FIG. 1 is a schematic diagram showing the structure of an on-board autonomous health management system according to an embodiment of the present invention. Figure 5As shown, an on-board autonomous health management system includes a data acquisition module 501, a micro-vibration source separation module 502 and a fault prediction module 503. The data acquisition module 501 is used to collect micro-vibration data during the on-orbit operation of the spacecraft, and the micro-vibration data includes normal single-machine vibration, abnormal single-machine vibration, normal space torque and whole-satellite vibration caused by abnormal space torque. The micro-vibration source separation module 502 is communicatively connected to the data acquisition module 501, and is used to separate the micro-vibration source based on the micro-vibration data, obtain micro-vibration source data, and send it to the fault prediction module 503. The fault prediction module 503 is used to perform vibration rationality prediction based on the micro-vibration source data and predict the fault state of the spacecraft. In one embodiment of the present invention, Figure 5 As shown, the prediction results of the fault prediction module 503 can be used as a basis for fault judgment of the autonomous health management on board, and can be combined with the autonomous health management system of the spacecraft for troubleshooting. Specifically, if abnormal sensor data appears in the spacecraft, the separation algorithm as described above can also be used to analyze the problem unit to verify the working status of the unit, thereby ensuring the stable operation of the spacecraft. On the other hand, the prediction results can also be sent to the ground measurement and control station. The ground measurement and control station and the ground system can realize comprehensive monitoring and analysis of the spacecraft's operating status by processing and analyzing multi-dimensional data, taking into account micro-vibration data, and combining other sensor data such as temperature and pressure, thereby improving the accuracy and comprehensiveness of fault prediction. In addition, as mentioned above, the micro-vibration data collected by the data acquisition module 501 can also be sent directly to the ground measurement and control station for training and verification of the sharp model in the separation method on the ground.

[0065] The present invention provides a method for separating spacecraft microvibration sources and a fault prediction method, which achieves precise analysis of spacecraft microvibrations on-orbit through a special microvibration interference source separation algorithm. The separation method can effectively separate different microvibration sources and microvibration intensities, thereby accurately determining the impact of microvibrations on the spacecraft. At the same time, the fault prediction method can predict faults of individual spacecraft and platforms based on historical data or perform corroborative analysis of telemetry anomalies, providing reliable data support capabilities. The fault prediction method can improve the accuracy and efficiency of spacecraft fault prediction and provide strong technical support for the safe operation of spacecraft.

[0066] Although various embodiments of the present invention have been described above, it should be understood that they are presented by way of example only and not limitation. It will be apparent to those skilled in the relevant art that various combinations, modifications, and variations may be made thereto without departing from the spirit and scope of the present invention. Therefore, the breadth and scope of the present invention disclosed herein should not be limited by the exemplary embodiments disclosed above, but should be defined solely in accordance with the appended claims and their equivalents.

Claims

1. A method for separating a spacecraft micro-vibration source, characterized in that: include: Extract features from the micro-vibration data of the spacecraft during on-orbit operation and obtain a feature map; Assigning a weight value to each channel of the feature map to obtain a feature vector; as well as The feature vector is mapped into a probability distribution to achieve separation and category matching of micro-vibration sources.

2. The separation method according to claim 1, wherein Before feature extraction, the micro-vibration data is preprocessed, wherein the preprocessing includes denoising and data format standardization.

3. The separation method according to claim 1, wherein Feature extraction is performed through filters composed of multiple convolutional layers.

4. The separation method according to claim 1, wherein A weight value is assigned to each channel of the feature map based on an attention mechanism.

5. The separation method according to claim 1, wherein The feature vector is mapped to a probability distribution through the Softmax layer.

6. A spacecraft fault prediction method based on deep learning, characterized in that: include: Collect micro-vibration data of spacecraft during on-orbit operation; Separating the micro-vibration data by the separation method according to any one of claims 1 to 5 to obtain micro-vibration source data; as well as The fault state of the spacecraft is predicted based on the micro-vibration source data through a fault prediction model, wherein the fault prediction model is deployed on the spacecraft.

7. The spacecraft fault prediction method according to claim 6, wherein: The fault prediction model is constructed through the following steps: Evaluate the impact of different micro-vibration sources on spacecraft performance based on micro-vibration source data; and Through deep learning technology, a fault prediction model is built based on the analysis of evaluation results and historical fault data.

8. The spacecraft fault prediction method according to claim 6, wherein: Also includes: The fault prediction model is trained and optimized based on historical micro-vibration data.

9. The spacecraft fault prediction method according to claim 6, wherein: Also includes: Sending the predicted fault status to the ground station; as well as The ground station analyzes the operating status of the spacecraft based on the data from the onboard temperature sensor and pressure sensor and the fault status.

10. An autonomous health management system on board a satellite, characterized in that: include: a data acquisition module configured to collect micro-vibration data of the spacecraft during its on-orbit operation; a micro-vibration source separation module, which is communicatively connected to the data acquisition module and is configured to separate the micro-vibration source based on the micro-vibration data to obtain micro-vibration source data; as well as A fault prediction module is communicatively connected to the micro-vibration source separation module and is configured to perform vibration rationality prediction based on the micro-vibration source data and predict the fault state of the spacecraft.

Images