A method for index characterization and evaluation based on aerospace equipment models

Abstract

This invention discloses a method for index representation and evaluation based on aerospace equipment models, comprising the following steps: S1, extracting a set of mission indicators and mapping them to each subsystem; S2, generating mission context data using generative adversarial networks (GANs); S3, extracting features through self-supervised learning to generate a dataset; S4, fusing data through multimodal learning to generate a unified feature representation; S5, constructing an index relationship graph between subsystems using graph neural networks (PNNs); S6, performing temporal modeling and mission progress prediction using temporal convolutional networks (TCNNs); S7, generating adversarial self-supervised learning-enhanced data and performing label self-calibration; S8, analyzing data through unsupervised learning and dynamically adjusting evaluation criteria. This invention, by combining GANs, self-supervised learning, graph neural networks, and temporal convolutional networks, achieves diversified data generation, subsystem relationship modeling, temporal prediction, and dynamic mission optimization for aerospace missions.

Description

Technical Field

[0001] This invention relates to the field of aerospace technology, and in particular to an index characterization and evaluation method based on aerospace equipment models. Background Technology

[0002] With the increasing complexity of space missions and the diversification of execution environments, improving the mission execution efficiency of space equipment and assessing potential anomalies during mission execution have become critical issues that urgently need to be addressed. Traditional space mission evaluation methods typically rely on expert experience and static models, lacking the ability to dynamically adjust, especially in reflecting the real-time interactions between subsystems and their impact on the overall mission. Furthermore, as missions become increasingly complex, the mission indicators and dependencies between subsystems become more intricate, rendering traditional rule-based evaluation methods often inadequate to handle these complex changes and potential mission risks.

[0003] Existing space mission evaluation systems generally rely on static mathematical models or expert-designed rule sets. These methods often lack sufficient accuracy and adaptability when facing complex and ever-changing mission scenarios. Especially in collaborative execution environments involving multiple missions and subsystems, mission evaluation systems cannot respond promptly to changes in mission status, resulting in the inability to provide early warnings of potential risks during mission execution. Furthermore, existing evaluation methods typically rely on a single data source, such as mission logs or sensor data, neglecting the effective fusion of multimodal data from multiple sensors and subsystems. This makes traditional mission evaluation methods lack comprehensiveness and accuracy, failing to effectively capture dynamic changes during mission execution.

[0004] Meanwhile, with the development of artificial intelligence technology, utilizing machine learning and deep learning methods to improve the evaluation and optimization of space missions has become a trend. Current research has begun to explore the use of deep learning methods to automatically extract key features from mission data and make predictions, but these methods still have some shortcomings. First, existing machine learning methods often face problems such as difficulty in data fusion and complex feature selection when dealing with high-dimensional, multimodal data, resulting in insufficient accuracy and real-time performance of mission evaluation results. Second, existing deep learning models usually rely on large amounts of labeled data for training, while labeling mission data is often difficult and costly, especially since much data is unlabeled during space mission execution. This prevents existing self-supervised learning methods from being fully applied to the processing of space mission data.

[0005] Therefore, how to provide an index characterization and evaluation method based on aerospace equipment models is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose an indicator representation and evaluation method based on aerospace equipment models. This invention combines advanced technologies such as generative adversarial networks, self-supervised learning, graph neural networks, and temporal convolutional networks to propose a novel task evaluation method. This method can generate diverse task simulation data and extract effective features from it, reducing reliance on manually labeled data and improving the diversity and quality of task datasets. Simultaneously, by using graph neural networks to establish indicator relationship graphs between various subsystems of aerospace equipment, it can accurately capture the dependencies and synergies between subsystems, thereby optimizing the overall task evaluation. By modeling the temporal data during task execution using temporal convolutional networks, it can predict task progress and the state changes of each subsystem in real time, and dynamically adjust task evaluation standards and execution strategies based on the actual progress of the task.

[0007] An index characterization and evaluation method based on an aerospace equipment model according to an embodiment of the present invention includes the following steps:

[0008] S1. Through mission requirement analysis, identify and decompose various performance indicators of the space mission, map these indicators to different subsystems of the space equipment, and generate a set of mission indicators.

[0009] S2. Based on generative adversarial networks, task context data is generated according to the task's indicator set to form task simulation data;

[0010] S3. Utilize self-supervised learning to automatically extract effective features from task simulation data and generate a dataset for training.

[0011] S4. Combine multimodal learning methods to fuse the dataset with data from different sensors to generate a unified feature representation;

[0012] S5. Based on the generated unified feature representation, a graph neural network is used to construct an index relationship graph between various subsystems of aerospace equipment, thereby optimizing the interdependencies between the subsystems.

[0013] S6. Apply a temporal convolutional network to perform temporal modeling of the index relationship graph, generate a temporal model, and analyze data on mission progress, spacecraft status, and environmental changes.

[0014] S7. Based on task simulation data, generate diverse data using generative adversarial self-supervised learning, and perform label self-labeling through self-supervised learning to provide diverse data for training time series models.

[0015] S8. Use unsupervised learning methods to analyze diverse data, automatically identify potential patterns and anomalies, and dynamically adjust evaluation criteria based on different stages of the task and changes in the environment.

[0016] Optionally, S2 specifically includes:

[0017] S21. Based on the task-specific indicator set, extract the core indicators of task execution from the task requirement analysis, and use the generator in the generative adversarial network to generate task context data.

[0018] S22. Construct a generative adversarial network discriminator to evaluate the authenticity of task context data, optimize the generator, and make the task context data more in line with the indicator set requirements of the task.

[0019] S23. In each training iteration, the generator generates task simulation data based on the task's metric set, input random noise vector, and high-dimensional feature representation.

[0020] S24. Optimize the generator through backpropagation based on the loss function output by the discriminator:

[0021]

[0022] Where L is the quantization formula, and D real For real-world task data, D fake For the generated task simulation data, D(·) is the discriminator's score on the realism of the task scenario;

[0023] S25. During the iteration process, the generator continuously adjusts its parameters to generate task context data with a set of metrics that better suit the task.

[0024] Optionally, S3 specifically includes:

[0025] S31. Extract key features from the generated task simulation dataset, and use a self-supervised learning algorithm to perform unsupervised feature learning on the data to generate preliminary feature representations.

[0026] S32. Use convolutional neural networks to process the temporal features of the task simulation data, generate high-dimensional feature representations, and capture the temporal dependencies and dynamic changes during task execution.

[0027] S33. Fuse the features extracted by self-supervised learning and convolutional neural networks to generate a dataset;

[0028] S34. Perform cluster analysis on the dataset, classify the data according to the task execution context, and generate labels for each task context.

[0029] S35. Use a deep autoencoder to reduce the dimensionality of the dataset and generate low-dimensional task data;

[0030] S36. Standardize low-dimensional task data:

[0031]

[0032] Where D standardized For standardization, D reduced The final dataset is generated from low-dimensional task data, where μ is the mean of the dataset and σ is the standard deviation.

[0033] Optionally, S4 specifically includes:

[0034] S41. Based on the generated dataset and multimodal datasets from different sensors, align the task simulation data with the sensor data, construct a unified input feature matrix, and perform synchronous processing on different data sources.

[0035] S42. Use a convolutional neural network to extract features from the multimodal dataset to obtain high-dimensional features for each modality;

[0036] S43. Use multimodal learning algorithms to fuse datasets and high-dimensional features to generate a unified feature representation. Use a weighted fusion strategy to represent multi-source data as a unified multi-dimensional vector.

[0037] S44. Weighted fusion of multi-dimensional vectors:

[0038]

[0039] Where F fused To fuse the feature matrix, n represents multiple dimensions, w i These are the weighting coefficients of the modal data features. High-dimensional features;

[0040] S45. Perform batch normalization on the fused feature matrix:

[0041]

[0042] Where F fused To fuse the feature matrix, F normalized Let μ be the normalized feature matrix, μ be the mean of the feature matrix, and σ be the standard deviation.

[0043] S46. Generate the final unified feature representation based on the normalized feature matrix.

[0044] Optionally, S5 specifically includes:

[0045] S51. Based on the generated unified feature representation, extract the task indicators of each subsystem from the task simulation data and generate the subsystem feature vector.

[0046] S52. Using the generated subsystem feature vectors and the index dependencies of the task, construct an index relationship diagram between the subsystems of the aerospace equipment;

[0047] S53. Using graph neural networks, index relationship graphs and subsystem feature vectors are used to update the features of each subsystem, capture the dependencies between subsystems, and generate updated feature representations for each subsystem.

[0048] S54. Through the information transmission mechanism, information is transmitted and aggregated between the nodes of each subsystem to generate the final subsystem feature representation;

[0049] S55. Based on the final subsystem feature representation, calculate the overall evaluation index of the task to provide a comprehensive evaluation result for task optimization.

[0050] Optionally, S6 specifically includes:

[0051] S61. Based on the final subsystem feature representation, the features of each subsystem are used as inputs to the time series data to construct a time series input matrix;

[0052] S62. Apply a temporal convolutional network to process the temporal input matrix, extract temporal features through multi-layer convolutional operations, and generate a deep feature representation of the temporal data;

[0053] S63. By using the causal convolution operation in the convolutional neural network, a temporal feature representation after causal convolution is generated to further optimize the temporal modeling;

[0054] H causal =σ(R) causal ·X time ·W causal );

[0055] Among them, H causal R represents the temporal features output by the causal convolution operation. causal X is the causal mask in causal convolution. time W is the time-series input matrix. causal Let σ be the weight matrix of the causal convolution, and σ be the activation function.

[0056] S64. Using the skip connection technique in the convolutional neural network, the outputs of each convolutional layer are merged to generate the final temporal feature representation.

[0057] S65. Based on the final temporal feature representation, predict the future state of task execution and generate the temporal prediction result of the task;

[0058] S66. Based on the time series prediction results, combined with the task objectives and constraints, dynamically adjust and optimize the task progress, rationally allocate task resources, and finally output the optimized task execution plan.

[0059] Optionally, S8 specifically includes:

[0060] S81. Based on the generated diverse data, unsupervised learning algorithms are used to perform cluster analysis on the diverse data to identify patterns in the task context;

[0061] S82. Use the K-means clustering algorithm to cluster diverse data and generate cluster labels c for each data point. i The clustering effect in different task scenarios is measured by calculating the density of points within the cluster.

[0062] S83. Based on the clustering effect, calculate the density distribution of each task context cluster, and use the density value to identify anomalies in diverse data.

[0063] S84. Analyze clustering results and outlier data, automatically adjust task evaluation criteria, and generate dynamically adjusted task evaluation criteria based on different stages of task execution and environmental changes by dynamically adjusting various performance indicators of the task.

[0064] S85. Optimize task progress, subsystem status, and resource allocation based on the dynamically adjusted task evaluation criteria:

[0065]

[0066] Among them, I final For the final task evaluation result, p represents the dimensions of the task evaluation criteria, and w represents the dimensions of the task evaluation criteria. i The weighting coefficients for the evaluation criteria, The revised task evaluation criteria;

[0067] S86. Monitor task execution through a real-time feedback mechanism, and adjust task evaluation criteria in real time by combining data from smart wearable devices and sensors.

[0068] The beneficial effects of this invention are:

[0069] This invention, by combining advanced technologies such as generative adversarial networks (GANs), self-supervised learning, graph neural networks, and temporal convolutional networks, successfully addresses many problems existing in current space mission evaluation methods, significantly improving the accuracy, real-time performance, and robustness of space mission evaluation. Firstly, by using GANs, this invention can generate diverse mission simulation data, greatly enriching the mission dataset and reducing reliance on manually labeled data. This not only increases data diversity but also ensures that the evaluation process covers more mission execution scenarios, including normal missions and potential failure modes, thus providing more comprehensive and accurate data support for mission evaluation.

[0070] Through self-supervised learning, this invention further improves the quality of the task dataset, automatically extracting effective features from the generated task simulation data, reducing reliance on labeled data, and lowering the cost and time of data labeling. Particularly in the task data generation process, this invention combines generative adversarial self-supervised learning, resulting in task data that is not only diverse but also self-labeled, ensuring the authenticity and validity of the data. This innovative design provides more reliable training data for subsequent task evaluation, further improving the accuracy of the evaluation system.

[0071] This invention also successfully solves the problem of modeling complex dependencies between subsystems using graph neural networks. The index relationship graph constructed using GCN can capture the interdependencies and synergies between various subsystems of aerospace equipment, thus accurately reflecting the impact of different subsystems on mission execution. This graph neural network-based modeling method enables mission evaluation to consider not only the performance of a single subsystem but also the comprehensive evaluation of the collaborative effects of all subsystems, ensuring the optimization of overall mission performance and various mission objectives.

[0072] Furthermore, this invention models temporal data during task execution using temporal convolutional networks, addressing the shortcomings of existing methods in task progress prediction and dynamic adjustment. Temporal convolutional networks can capture temporal dependencies during task execution, accurately predict key indicators such as task progress, success probability, and resource consumption, and dynamically adjust evaluation criteria and execution strategies based on the actual progress of the task. This dynamic evaluation method based on temporal modeling ensures that the task can respond to changes in its state in real time, further enhancing the flexibility and robustness of task execution.

[0073] By employing multimodal learning, this invention addresses the challenge of multimodal data fusion in existing technologies. By effectively fusing data from different sensors, a unified feature representation is generated, providing more comprehensive and accurate data support for task evaluation. This data fusion technology significantly improves data utilization efficiency during task evaluation, ensuring the accuracy and stability of the evaluation system when handling complex tasks and multimodal data. Attached Figure Description

[0074] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0075] Figure 1 This is a flowchart of an index characterization and evaluation method based on aerospace equipment models proposed in this invention;

[0076] Figure 2This is a detailed structural diagram of the time-series modeling and dynamic prediction of an index characterization and evaluation method based on aerospace equipment models proposed in this invention. Detailed Implementation

[0077] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0078] refer to Figure 1-2 A method for index characterization and evaluation based on aerospace equipment models includes the following steps:

[0079] S1. Through mission requirement analysis, identify and decompose various performance indicators of the space mission, map these indicators to different subsystems of the space equipment, and generate a set of mission indicators.

[0080] S2. Based on generative adversarial networks, task context data is generated according to the task's indicator set to form task simulation data;

[0081] S3. Utilize self-supervised learning to automatically extract effective features from task simulation data and generate a dataset for training.

[0082] S4. Combine multimodal learning methods to fuse the dataset with data from different sensors to generate a unified feature representation;

[0083] S5. Based on the generated unified feature representation, a graph neural network is used to construct an index relationship graph between various subsystems of aerospace equipment, thereby optimizing the interdependencies between the subsystems.

[0084] S6. Apply a temporal convolutional network to perform temporal modeling of the index relationship graph, generate a temporal model, and analyze data on mission progress, spacecraft status, and environmental changes.

[0085] S7. Based on task simulation data, generate diverse data using generative adversarial self-supervised learning, and perform label self-labeling through self-supervised learning to provide diverse data for training time series models.

[0086] S8. Use unsupervised learning methods to analyze diverse data, automatically identify potential patterns and anomalies, and dynamically adjust evaluation criteria based on different stages of the task and changes in the environment.

[0087] In this embodiment, S2 specifically includes:

[0088] S21. Based on the task-specific indicator set, extract the core indicators of task execution from the task requirement analysis, and use the generator in the generative adversarial network to generate task context data.

[0089] S22. Construct a generative adversarial network discriminator to evaluate the authenticity of task context data, optimize the generator, and make the task context data more in line with the indicator set requirements of the task.

[0090] S23. In each training iteration, the generator generates task simulation data based on the task's metric set, input random noise vector, and high-dimensional feature representation.

[0091] S24. Optimize the generator through backpropagation based on the loss function output by the discriminator:

[0092]

[0093] Where L is the quantization formula, and D real For real-world task data, D fake For the generated task simulation data, D(·) is the discriminator's score on the realism of the task scenario;

[0094] S25. During the iteration process, the generator continuously adjusts its parameters to generate task context data with a set of metrics that better suit the task.

[0095] In this embodiment, S3 specifically includes:

[0096] S31. Extract key features from the generated task simulation dataset, and use a self-supervised learning algorithm to perform unsupervised feature learning on the data to generate preliminary feature representations.

[0097] S32. Use convolutional neural networks to process the temporal features of the task simulation data, generate high-dimensional feature representations, and capture the temporal dependencies and dynamic changes during task execution.

[0098] S33. Fuse the features extracted by self-supervised learning and convolutional neural networks to generate a dataset;

[0099] S34. Perform cluster analysis on the dataset, classify the data according to the task execution context, and generate labels for each task context.

[0100] S35. Use a deep autoencoder to reduce the dimensionality of the dataset and generate low-dimensional task data;

[0101] S36. Standardize low-dimensional task data:

[0102]

[0103] Where D standardized For standardization, D reduced The final dataset is generated from low-dimensional task data, where μ is the mean of the dataset and σ is the standard deviation.

[0104] In this embodiment, S4 specifically includes:

[0105] S41. Based on the generated dataset and multimodal datasets from different sensors, align the task simulation data with the sensor data, construct a unified input feature matrix, and perform synchronous processing on different data sources.

[0106] S42. Use a convolutional neural network to extract features from the multimodal dataset to obtain high-dimensional features for each modality;

[0107] S43. Use multimodal learning algorithms to fuse datasets and high-dimensional features to generate a unified feature representation. Use a weighted fusion strategy to represent multi-source data as a unified multi-dimensional vector.

[0108] S44. Weighted fusion of multi-dimensional vectors:

[0109]

[0110] Where F fused To fuse the feature matrix, n represents multiple dimensions, w i These are the weighting coefficients of the modal data features. High-dimensional features;

[0111] S45. Perform batch normalization on the fused feature matrix:

[0112]

[0113] Where F fused To fuse the feature matrix, F normalized Let μ be the normalized feature matrix, μ be the mean of the feature matrix, and σ be the standard deviation.

[0114] S46. Generate the final unified feature representation based on the normalized feature matrix.

[0115] In this embodiment, S5 specifically includes:

[0116] S51. Based on the generated unified feature representation, extract the task indicators of each subsystem from the task simulation data and generate the subsystem feature vector.

[0117] S52. Using the generated subsystem feature vectors and the index dependencies of the task, construct an index relationship diagram between the subsystems of the aerospace equipment;

[0118] S53. Using graph neural networks, index relationship graphs and subsystem feature vectors are used to update the features of each subsystem, capture the dependencies between subsystems, and generate updated feature representations for each subsystem.

[0119] S54. Through the information transmission mechanism, information is transmitted and aggregated between the nodes of each subsystem to generate the final subsystem feature representation;

[0120] S55. Based on the final subsystem feature representation, calculate the overall evaluation index of the task to provide a comprehensive evaluation result for task optimization.

[0121] In this embodiment, S6 specifically includes:

[0122] S61. Based on the final subsystem feature representation, the features of each subsystem are used as inputs to the time series data to construct a time series input matrix;

[0123] S62. Apply a temporal convolutional network to process the temporal input matrix, extract temporal features through multi-layer convolutional operations, and generate a deep feature representation of the temporal data;

[0124] S63. By using the causal convolution operation in the convolutional neural network, a temporal feature representation after causal convolution is generated to further optimize the temporal modeling;

[0125] H causal =σ(R) causal ·X time ·W causal );

[0126] Among them, H causal R represents the temporal features output by the causal convolution operation. causal X is the causal mask in causal convolution. time W is the time-series input matrix. causal Let σ be the weight matrix of the causal convolution, and σ be the activation function.

[0127] S64. Using the skip connection technique in the convolutional neural network, the outputs of each convolutional layer are merged to generate the final temporal feature representation.

[0128] S65. Based on the final temporal feature representation, predict the future state of task execution and generate the temporal prediction result of the task;

[0129] S66. Based on the time series prediction results, combined with the task objectives and constraints, dynamically adjust and optimize the task progress, rationally allocate task resources, and finally output the optimized task execution plan.

[0130] In this embodiment, S8 specifically includes:

[0131] S81. Based on the generated diverse data, unsupervised learning algorithms are used to perform cluster analysis on the diverse data to identify patterns in the task context;

[0132] S82. Use the K-means clustering algorithm to cluster diverse data and generate cluster labels c for each data point. i The clustering effect in different task scenarios is measured by calculating the density of points within the cluster.

[0133] S83. Based on the clustering effect, calculate the density distribution of each task context cluster, and use the density value to identify anomalies in diverse data.

[0134] S84. Analyze clustering results and outlier data, automatically adjust task evaluation criteria, and generate dynamically adjusted task evaluation criteria based on different stages of task execution and environmental changes by dynamically adjusting various performance indicators of the task.

[0135] S85. Optimize task progress, subsystem status, and resource allocation based on the dynamically adjusted task evaluation criteria:

[0136]

[0137] Among them, I final For the final task evaluation result, p represents the dimensions of the task evaluation criteria, and w represents the dimensions of the task evaluation criteria. i The weighting coefficients for the evaluation criteria, The revised task evaluation criteria;

[0138] S86. Monitor task execution through a real-time feedback mechanism, and adjust task evaluation criteria in real time by combining data from smart wearable devices and sensors.

[0139] Example 1:

[0140] To verify the feasibility of this invention in practice, it was applied to a simulated deep space exploration mission. The mission objective was to simulate a spacecraft performing a series of complex operations during exploration, including planetary orbit adjustments, data acquisition, and energy management. The spacecraft used in this mission comprises multiple subsystems, such as propulsion, energy management, thermal control, and communication systems, each playing a crucial role in mission execution. Because missions are subject to various environmental factors over extended periods, and complex dependencies exist between the subsystems, traditional mission evaluation methods often fail to effectively predict mission progress, resource consumption, and potential failures.

[0141] In this simulation task, the method of this invention is applied to address the problems of existing evaluation methods, such as the inability to respond in real time to dynamic changes in task execution, scarcity of task data, and insufficient modeling of dependencies between subsystems. By generating diverse task simulation data through generative adversarial networks, the task evaluation system can simulate the task execution process under different scenarios, such as normal tasks, propulsion system failures, and energy shortages, greatly enhancing the diversity and coverage of task data.

[0142] In the initial stage of the mission, 50,000 simulation data points were generated based on a generative adversarial network (GAN), including normal execution scenarios, propulsion system failures, communication interruptions, and abnormal energy consumption. This data not only covers normal situations that may occur during mission execution but also simulates sudden failures and mission execution scenarios under complex environments. Through self-supervised learning, the system automatically extracted key features from these generated mission data, such as mission success rate, propulsion efficiency, communication latency, and temperature fluctuations, providing rich feature data for subsequent mission evaluation.

[0143] Subsequently, a graph neural network was used to construct an index relationship diagram between the spacecraft's subsystems, capturing the interdependencies between them. Through the graph neural network model, the system can adjust the operational status of each subsystem in real time to ensure the smooth execution of the mission. When the propulsion system experiences performance degradation, the graph neural network can promptly identify and prompt the energy management system to reallocate resources, thereby avoiding potential energy shortages during the mission.

[0144] During mission execution, temporal convolutional networks (CCNNs) were used to model and predict the temporal data of the mission. By performing temporal modeling on real-time data of each subsystem of the spacecraft during the mission, CCNNs were able to accurately predict the mission's progress and the state changes of each subsystem. At the 100th hour of the mission, the CCNN predicted that the propulsion system's fuel consumption was about to exceed the predetermined value, potentially causing the mission to fail to complete on time. Based on this prediction, the mission evaluation system adjusted its energy allocation strategy, prioritizing the propulsion system's energy needs, successfully avoiding fuel shortages and ensuring the smooth progress of the mission in subsequent phases.

[0145] After the task was completed, the system compared the performance of traditional task evaluation methods with that of the method of this invention. Traditional methods, based on static models and manually defined rules, achieved a success rate of 85%, while the method of this invention, through real-time prediction and dynamic adjustment of task evaluation criteria, achieved a final success rate of 95%. Regarding fault prediction, the traditional method achieved an accuracy rate of 70%, while the method of this invention, through dynamic evaluation combining generative adversarial networks and graph neural networks with temporal convolutional networks, improved fault prediction accuracy to 92%. In terms of resource utilization, the traditional method achieved a resource utilization rate of 80%, while the method of this invention, through dynamic task adjustment and resource optimization, increased resource utilization to 90%.

[0146] Table 1 Statistics of the Task Simulation Dataset

[0147]

[0148] Table 2 Comparison of Task Execution Performance

[0149]

[0150] Table 1 illustrates the basic information of the mission simulation dataset generated during the implementation of this invention. The dataset contains 50,000 mission simulation data points, covering normal, fault, and special mission scenarios, ensuring the comprehensiveness and diversity of mission evaluation. The generated mission data covers over 95% of mission execution scenarios, including normal mission execution, propulsion system failure, communication interruption, and abnormal energy consumption. As shown in the table, normal mission data accounts for 60% of the mission simulation data. This type of data reflects the spacecraft's mission execution under normal conditions, including the standard operating states of various subsystems such as the propulsion system, energy system, and communication system. Fault mission data and special scenario mission data account for 30% and 10%, respectively. These data simulate possible faults or abnormal situations during the mission, such as propulsion system failure, energy system anomalies, and communication interruptions. These fault and abnormal scenario data are crucial for assessing potential risks and problems during the mission, helping the mission evaluation model better cope with complex and ever-changing real-world mission scenarios.

[0151] Table 2 provides a comparison of traditional task evaluation methods and the method of this invention across different evaluation metrics. By comparing multiple task evaluation metrics, the significant advantages of the method of this invention over the traditional method are demonstrated in terms of task success rate, fault prediction accuracy, resource utilization, and task completion time. The traditional method achieves 85%, while the method of this invention reaches 95%, an improvement of 10%. The accuracy rate of the traditional method is 70%, while the accuracy rate of the method of this invention is 92%, an improvement of 22%. The resource utilization rate of the traditional method is 80%, while the resource utilization rate of the method of this invention is 90%, an improvement of 10%. The traditional method requires 120 hours to complete the task, while the method of this invention can shorten the task completion time to 108 hours, a reduction of 10%.

[0152] The data in Tables 1 and 2 demonstrate that the method of this invention has significant advantages over traditional methods in several aspects. First, the diverse task data generated by generative adversarial networks allows task evaluation to cover more task scenarios, reducing the problems of data scarcity and insufficient coverage in traditional methods. Second, graph neural networks can accurately model the dependencies between various subsystems of the spacecraft, providing more accurate subsystem state predictions for task evaluation. The application of temporal convolutional networks enables real-time prediction and dynamic adjustment of key indicators such as task progress and resource consumption, avoiding time delays and resource waste caused by inaccurate task execution in traditional methods.

[0153] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for index characterization and evaluation based on aerospace equipment models, characterized in that, Includes the following steps: S1. Through mission requirement analysis, identify and decompose various performance indicators of the space mission, map various performance indicators to different subsystems of the space equipment, and generate a set of mission indicators. S2. Based on generative adversarial networks, task context data is generated according to the task's indicator set to form task simulation data; S3. Utilize self-supervised learning to automatically extract effective features from task simulation data and generate a dataset for training. S4. Combine multimodal learning methods to fuse the dataset with data from different sensors to generate a unified feature representation; S5. Based on the generated unified feature representation, a graph neural network is used to construct an index relationship graph between various subsystems of aerospace equipment, thereby optimizing the interdependencies between the subsystems. S6. Apply a temporal convolutional network to perform temporal modeling of the index relationship graph, generate a temporal model, and analyze data on mission progress, spacecraft status, and environmental changes. S7. Based on task simulation data, generate diverse data using generative adversarial self-supervised learning, and perform label self-labeling through self-supervised learning to provide diverse data for training time series models. S8. Use unsupervised learning methods to analyze diverse data, automatically identify potential patterns and anomalies, and dynamically adjust evaluation criteria based on different stages of the task and changes in the environment.

2. The index characterization and evaluation method based on aerospace equipment models according to claim 1, characterized in that, S2 specifically includes: S21. Based on the task-specific indicator set, extract the core indicators of task execution from the task requirement analysis, and use the generator in the generative adversarial network to generate task context data. S22. Construct a generative adversarial network discriminator to evaluate the authenticity of task context data, optimize the generator, and make the task context data more in line with the indicator set requirements of the task. S23. In each training iteration, the generator generates task simulation data based on the task's metric set, input random noise vector, and high-dimensional feature representation. S24. Optimize the generator through backpropagation based on the loss function output by the discriminator: ； in, For quantification formula, This is real task data. For the generated task simulation data, The discriminator outputs a score indicating the realism of the task context; S25. During the iteration process, the generator continuously adjusts its parameters to generate task context data with a set of metrics that better suit the task.

3. The index characterization and evaluation method based on aerospace equipment models according to claim 1, characterized in that, S3 specifically includes: S31. Extract key features from the generated task simulation data, and use a self-supervised learning algorithm to perform unsupervised feature learning on the task simulation data to generate preliminary feature representations. S32. Use convolutional neural networks to process the temporal features of the task simulation data, generate high-dimensional feature representations, and capture the temporal dependencies and dynamic changes during task execution. S33. Fuse the features extracted by self-supervised learning and convolutional neural networks to generate a dataset; S34. Perform cluster analysis on the dataset, classify the data according to the task execution context, and generate labels for each task context. S35. Use a deep autoencoder to reduce the dimensionality of the dataset and generate low-dimensional task data; S36. Standardize low-dimensional task data: ； in For standardization, For low-dimensional task data, The mean of the dataset. The standard deviation is used to generate the final dataset.

4. The index characterization and evaluation method based on aerospace equipment models according to claim 1, characterized in that, S4 specifically includes: S41. Based on the generated dataset and multimodal datasets from different sensors, align the task simulation data with the sensor data, construct a unified input feature matrix, and perform synchronous processing on different data sources. S42. Use a convolutional neural network to extract features from the multimodal dataset to obtain high-dimensional features for each modality; S43. Use multimodal learning algorithms to fuse datasets and high-dimensional features to generate a unified feature representation. Use a weighted fusion strategy to represent multi-source data as a unified multi-dimensional vector. S44. Weighted fusion of multi-dimensional vectors: ； in To fuse the feature matrix, n represents multiple dimensions. These are the weighting coefficients of the modal data features. High-dimensional features; S45. Perform batch normalization on the fused feature matrix: ； in To fuse the feature matrix, The normalized feature matrix, The mean of the characteristic matrix is... Standard deviation; S46. Generate the final unified feature representation based on the normalized feature matrix.

5. The index characterization and evaluation method based on aerospace equipment models according to claim 1, characterized in that, S5 specifically includes: S51. Based on the generated unified feature representation, extract the task indicators of each subsystem from the task simulation data and generate the subsystem feature vector. S52. Using the generated subsystem feature vectors and the index dependencies of the task, construct an index relationship diagram between the subsystems of the aerospace equipment; S53. Using graph neural networks, index relationship graphs and subsystem feature vectors are used to update the features of each subsystem, capture the dependencies between subsystems, and generate updated feature representations for each subsystem. S54. Through the information transmission mechanism, information is transmitted and aggregated between the nodes of each subsystem to generate the final subsystem feature representation; S55. Based on the final subsystem feature representation, calculate the overall evaluation index of the task to provide a comprehensive evaluation result for task optimization.

6. The index characterization and evaluation method based on aerospace equipment models according to claim 5, characterized in that, S6 specifically includes: S61. Based on the final subsystem feature representation, the features of each subsystem are used as inputs to the time series data to construct a time series input matrix; S62. Apply a temporal convolutional network to process the temporal input matrix, extract temporal features through multi-layer convolutional operations, and generate a deep feature representation of the temporal data; S63. By using the causal convolution operation in the convolutional neural network, a temporal feature representation after causal convolution is generated to further optimize the temporal modeling; ； in, This represents the temporal features output by the causal convolution operation. For causal masks in causal convolution, For the time series input matrix, The weight matrix for causal convolution is... For activation functions; S64. Using the skip connection technique in the convolutional neural network, the outputs of each convolutional layer are merged to generate the final temporal feature representation. S65. Based on the final temporal feature representation, predict the future state of task execution and generate the temporal prediction result of the task; S66. Based on the time series prediction results, combined with the task objectives and constraints, dynamically adjust and optimize the task progress, rationally allocate task resources, and finally output the optimized task execution plan.

7. The index characterization and evaluation method based on aerospace equipment models according to claim 1, characterized in that, S8 specifically includes: S81. Based on the generated diverse data, unsupervised learning algorithms are used to perform cluster analysis on the diverse data to identify patterns in the task context; S82. Use the K-means clustering algorithm to cluster diverse data and generate cluster labels for each data point. The clustering effect in different task scenarios is measured by calculating the density of points within the cluster. S83. Based on the clustering effect, calculate the density distribution of each task context cluster, and use the density value to identify anomalies in diverse data. S84. Analyze clustering results and outlier data, automatically adjust task evaluation criteria, and generate dynamically adjusted task evaluation criteria based on different stages of task execution and environmental changes by dynamically adjusting various performance indicators of the task. S85. Optimize task progress, subsystem status, and resource allocation based on the dynamically adjusted task evaluation criteria: ； in, For the final task evaluation results, Dimensions of task evaluation criteria The weighting coefficients for the evaluation criteria, The revised task evaluation criteria; S86. Monitor task execution through a real-time feedback mechanism, and adjust task evaluation criteria in real time by combining data from smart wearable devices and sensors.

Images