Space station payload space-ground collaborative autonomous health management system

Abstract

The invention relates to the technical field of spacecrafts, in particular to a space station payload space-ground collaborative autonomous health management system, which is characterized in that remote measurement and work parameter data of load equipment at each level are acquired in real time through a data acquisition module, and health state information is formed; the fault characterization module analyzes the fault knowledge and generates a fault plan; the fault diagnosis module combines the pre-arranged plan and the health state information, analyzes a fault mode, outputs an abnormal result, gives an alarm and determines a fault type; the experiment control module is used for collecting experiment sample image video data and environment parameters and formulating an experiment regulation and control strategy; the life prediction module predicts the residual life of the equipment based on the fault plan and the health state information; and the scheme updating module integrates the fault diagnosis type, the experimental data and the life prediction result, and generates a target health management scheme. All the modules work cooperatively, the health management efficiency and autonomy of the effective load of the space station are improved, the dependence on ground support is reduced, and the reliability of long-term in-orbit operation is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of spacecraft, and in particular to a space station payload earth-space cooperative autonomous health management system. BACKGROUND

[0002] At present, the health management in the field of spaceflight at home and abroad generally adopts a strategy of being mainly passive and secondarily active, and is mainly based on ground remote flight control. The ground system receives and analyzes telemetry data in real time, and the operation and management personnel on the ground manually monitor and use automatic interpretation programs to monitor the state of the load, and cooperates with experts and scientists to carry out fault management and disposal. However, due to the discontinuity of space-ground measurement and control resources, the obstruction of data downlink, and the insufficient emergency response capability of the ground, the ground system cannot monitor and dispose the running state of the on-orbit load in real time all day long, and it is difficult to take fault isolation measures in time, and there is a risk of serious failure or accident.

[0003] With the continuous enhancement of the computing capacity of the space station, the support for on-orbit data processing, real-time fault reasoning, autonomous maintenance decision-making and other capabilities has also been greatly improved. By moving some health management business points with real-time sensing and emergency disposal needs to the edge of the space-based terminal, the on-orbit autonomous response and decision-making capability of the load health management can be improved. However, the space-based fully autonomous mode is limited by on-orbit computing and storage resources, and it is difficult to support the effective deployment and application of too complex deep learning algorithms.

[0004] Compared with the ground system mode and the space-based fully autonomous mode, the earth-space cooperative health management mode makes full use of the support capability of large-scale computing resources on the ground for complex model algorithms and the all-weather emergency response capability of space-based computing resources, and can effectively guarantee the immediacy and accuracy of the space station payload health management, and improve its autonomous ability of fault detection, diagnosis and disposal, experimental safety analysis and control, health assessment prediction and maintenance. SUMMARY

[0005] Therefore, the present application provides a space station payload earth-space cooperative autonomous health management system to solve the foregoing problems in the prior art.

[0006] To achieve the above-mentioned purpose, the present application provides a space station payload earth-space cooperative autonomous health management system, comprising:

[0007] A data acquisition module is used to acquire telemetry data and operating parameter data of all devices at the load level, subsystem level and system level in real time to obtain health state information;

[0008] A fault characterization module is used to acquire fault knowledge and analyze it to obtain a plurality of fault plans;

[0009] a fault diagnosis module connected with the data collection module and the fault characterization module, configured to analyze a fault mode according to the fault preplan and the health state information to obtain an abnormal result and alarm, and diagnose the abnormal result to obtain a fault diagnosis type;

[0010] an experiment control module configured to collect images and videos of an experimental sample to obtain health state data, collect environmental parameters and task safety requirements, and analyze the health state data according to the environmental parameters and the task safety requirements to obtain an experimental control strategy;

[0011] a life prediction module configured to analyze the fault preplan and the health state information to predict a remaining life of the equipment to obtain a prediction result;

[0012] a scheme updating module connected with the fault diagnosis module, the experiment control module and the life prediction module, configured to generate the initial health management scheme according to the fault diagnosis type, the experimental control strategy and the life prediction result, use a machine learning model to perform fault detection on an unknown fault according to the health management scheme and the health state information to obtain a detection result, and optimize the initial health management scheme according to the detection result to obtain a target health management scheme.

[0013] Further, the faults in the fault preplan can be divided into system-level faults, regular faults, experimental sample safety faults and degradation faults, the unknown fault is a preplan-outside fault, and the fault diagnosis type includes interface faults, function faults and performance degradation faults.

[0014] Further, the fault characterization module includes:

[0015] a knowledge collection unit configured to analyze interface data, fault mode influence analysis and a fault tree to obtain fault knowledge;

[0016] a relationship mapping unit connected with the knowledge collection unit, configured to define a parameter set related to a health state according to the fault knowledge to obtain a fault criterion;

[0017] a preplan generation unit connected with the knowledge collection unit and the relationship mapping unit respectively, configured to define fault response measures and operation processes according to the fault knowledge and the fault criterion, and integrate to form a plurality of fault preplans.

[0018] Further, the fault diagnosis module includes:

[0019] a data processing unit configured to perform feature extraction and analysis on the health state information to obtain a processing result;

[0020] A dynamic adaptation unit, connected with the data processing unit, is configured to dynamically adjust threshold rules and statistical rules according to the processing result to obtain an adjustment result to adapt to different environments.

[0021] An anomaly diagnosis unit, connected with the dynamic adaptation unit, is configured to perform matching analysis on a fault plan according to the processing result and the adjustment result to generate an anomaly result priority ranking, and perform probability diagnosis on the anomaly result by using a statistical analysis algorithm to obtain the fault diagnosis type.

[0022] Further, the anomaly diagnosis unit comprises:

[0023] A plan matching sub-unit is configured to perform matching analysis on a fault plan according to the adjustment result and the processing result to generate matching degree data.

[0024] A priority ranking sub-unit, connected with the plan matching sub-unit, is configured to perform priority ranking on the matching result to generate a ranking result based on the severity and occurrence probability of an anomaly property.

[0025] A probability diagnosis sub-unit, connected with the priority ranking sub-unit, is configured to perform probability evaluation on a fault diagnosis type according to the ranking result to obtain a probability distribution of each fault diagnosis type, and further determine the fault diagnosis type.

[0026] Further, the life prediction module comprises:

[0027] A feature extraction unit is configured to extract life-related features from the health state information to obtain an extraction result.

[0028] A degradation modeling unit, connected with the feature extraction unit, is configured to perform degradation modeling on key features in the extraction result to obtain a health state description of the device.

[0029] A life prediction unit, connected with the feature modeling unit, is configured to perform remaining life prediction according to the fault plan and the health state description to obtain the prediction result.

[0030] Further, the life prediction unit comprises:

[0031] A threshold determination sub-unit is configured to integrate failure thresholds of key features in the health state description and the fault plan to obtain an integration result.

[0032] A life prediction sub-unit, connected with the feature fusion sub-unit, is configured to perform preliminary remaining life prediction according to the integration result to obtain an initial prediction result.

[0033] A result optimization sub-unit, connected with the life prediction sub-unit, is configured to perform error analysis and optimization on the initial prediction result to obtain the prediction result.

[0034] Further, the scheme updating module comprises:

[0035] a scheme integration unit configured to generate an initial health management scheme comprising a fault handling strategy, a parameter control strategy and a maintenance and repair strategy according to the fault diagnosis type, the experimental safety risk and the life prediction result;

[0036] a fault detection unit connected with the scheme integration unit and configured to acquire historical health state information, and identify the unknown fault using a machine learning model according to the historical health state information to obtain a detection result;

[0037] a scheme updating unit connected with the fault detection unit and configured to feed back the initial health management scheme according to the detection result to obtain a target health management scheme.

[0038] Further, the fault detection unit comprises:

[0039] a model training subunit configured to, after data preprocessing of the historical health state information, use the historical health state information as a training set to train the machine learning model to obtain a trained model;

[0040] a feature engineering subunit configured to analyze the health state information to extract a feature vector for fault detection;

[0041] a result generation subunit connected with the feature engineering subunit and configured to use the feature vector as an input of the trained model to obtain the detection result.

[0042] Further, the space-based part transmits the collected health state information to the ground-based part, the ground-based part performs offline training and optimization of a fault characterization and diagnosis model based on the health state information and the historical health state information, and updates the optimized model to the space-based part through a space-ground link; the space-based part performs real-time fault detection, isolation and recovery on the health state of the payload according to the optimized model, and feeds back an execution result to the ground-based part; the ground-based part further adjusts and optimizes the model according to the execution result, to realize dynamic optimization of the space-ground cooperation.

[0043] Compared with the prior art, the beneficial effects of the present application are that the present application comprehensively monitors the health status of the equipment and the experimental sample situation through the cooperative work of the data acquisition module and the experimental control module, and provides accurate data support for fault diagnosis and life prediction. The close cooperation of the fault characterization module and the fault diagnosis module realizes rapid and accurate diagnosis and early warning of equipment failure, and ensures the reliable operation of the equipment. The scheme updating module generates and optimizes the health management scheme based on the fault diagnosis result, the experimental control strategy and the life prediction, and improves the intelligent management level of the system. The modules cooperate closely to realize the optimization of space and ground cooperation, significantly improve the accuracy and timeliness of fault diagnosis, enhance the autonomy and reliability of the system, prolong the service life of the equipment, reduce the operating cost, improve the resource utilization efficiency, and ensure the effective operation of the space station load.

[0044] Especially, based on the fault mode definition, the parameter set related to the health state is formed, the reliable fault criterion is formed, the accuracy of fault diagnosis is improved, and misjudgment and omission are avoided. A clear mapping relationship between the fault mode and the health state parameter is established, so that the system can quickly and accurately judge the fault condition according to the parameter change, and provide strong support for timely handling of faults. According to the fault knowledge and the fault criterion, detailed fault response measures and operation processes are defined to ensure that the fault can be quickly and effectively handled when it occurs, and the loss caused by the fault is reduced.

[0045] Especially, through feature extraction and modeling, the key features of the equipment health state can be accurately captured, thereby improving the accuracy of the remaining life prediction. Potential faults are discovered in time, and maintenance or replacement is performed in advance to reduce equipment downtime and maintenance costs. Reasonably arrange the maintenance plan, optimize the spare parts inventory, and reduce the operating cost. By predicting the remaining life of the equipment, measures can be taken in advance to avoid safety accidents caused by equipment failure.

[0046] Especially, by comprehensively considering the fault diagnosis type, the target health management data and the life prediction result, the generated health management scheme can more comprehensively cover various fault conditions that may occur in the equipment, and improve the accuracy and effectiveness of fault management. Unknown faults are detected using a machine learning model, and the health management scheme is optimized according to the detection result, so that the system can better adapt to the uncertainty and changes in the equipment operation process, and improve the flexibility and robustness of the system. Through accurate life prediction and fault detection, facilities maintenance and resource allocation can be more reasonably arranged to avoid unnecessary maintenance and resource waste, and reduce operating costs. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 A structural schematic diagram of a space station payload space-ground cooperative autonomous health management system provided by the present application is provided.

[0048] Figure 2A structural schematic diagram of the fault characterization module in the space station payload space-ground collaborative autonomous health management system provided by the application is shown in the figure;

[0049] Figure 3 A structural schematic diagram of the fault diagnosis module in the space station payload space-ground collaborative autonomous health management system provided by the application is shown in the figure;

[0050] Figure 4 A schematic diagram of the space-ground collaborative autonomous health management architecture provided by the application is shown in the figure;

[0051] Figure 5 A schematic diagram of the on-orbit distributed architecture of the autonomous health management provided by the application is shown in the figure. DETAILED DESCRIPTION

[0052] In order to make the objects and advantages of the present application clearer, the present application will be further described below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application.

[0053] The preferred embodiments of the present application will be described below with reference to the accompanying drawings. It should be understood by those skilled in the art that the embodiments are only used to explain the technical principles of the present application and are not used to limit the protection scope of the present application.

[0054] It should be noted that, in the description of the present application, the terms indicating the direction or position relationship such as "upper", "lower", "left", "right", "inner", "outer" and the like are based on the direction or position relationship shown in the drawings, which is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application.

[0055] In addition, it should also be noted that, in the description of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connecting", "connection" should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integrally connected; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or the internal communication of two elements. Those skilled in the art can understand the specific meaning of the above terms in the present application according to the specific circumstances.

[0056] Please refer to Figure 1 The space station payload space-ground collaborative autonomous health management system provided by the application is shown in the figure, which comprises:

[0057] The data acquisition module 10 is used to acquire the telemetry data and the working parameter data of all devices at the payload level, subsystem level and system level in real time to obtain the health state information;

[0058] Specifically, telemetry data (such as temperature, pressure, current, etc.) and operating parameter data (such as task period, execution frequency, etc.) are collected in real time from the device interfaces of the load level, subsystem level and system level. The collected raw data is subjected to noise filtering, missing value filling and normalization processing to ensure the accuracy and consistency of the data. The preprocessed data is used as health state information.

[0059] The fault characterization module 20 is configured to acquire fault knowledge and analyze the fault knowledge to obtain a plurality of fault plans.

[0060] Specifically, as shown in FIG. 1, the fault characterization module 20 includes: Figure 2

[0061] The knowledge acquisition unit 11 is configured to analyze the interface data sheet, fault mode influence analysis and fault tree to obtain fault knowledge.

[0062] Specifically, the device interface data sheet is collected, possible fault modes of the interface are identified, such as electrical interface short circuit, data interface signal error, etc., and the influence of the fault modes on the device function is analyzed. The device fault modes are classified, the influence degree of each fault mode is evaluated, and the key fault modes are identified. A fault tree model is constructed, the fault probability is calculated, and the key path leading to the device failure is found.

[0063] The relationship mapping unit 12 is connected with the knowledge acquisition unit 11 and is configured to define a set of parameters related to the health state according to the fault knowledge to obtain a fault criterion.

[0064] Specifically, the device health state parameters related to the fault modes are identified, and the change rule thereof is analyzed. The fault criterion of the health state parameters for each fault mode is set, such as the threshold range of the parameters, and a comprehensive criterion is established. The mapping relationship between the health state parameters and the fault modes is established by using a mathematical model or a data-driven method. The mathematical model can be based on the physical principles and working characteristics of the device, such as the heat balance equation of the motor, the measurement equation of the sensor, etc. The data-driven method can learn the internal relationship between the parameters and the fault modes from a large amount of historical health state information and health state data by using a machine learning algorithm (such as regression analysis, neural network, etc.).

[0065] The plan generation unit 13 is connected with the knowledge acquisition unit 11 and the relationship mapping unit 12, respectively, and is configured to define fault response measures and operation processes according to the fault knowledge and the fault criterion, and integrate the same to form a plurality of fault plans.

[0066] ​Specifically, countermeasures are formulated for each failure mode, and the measures are classified according to the degree of failure influence. The operation flow of failure diagnosis and handling is designed, and the operation steps, responsibility division and required resources are specified. The failure countermeasures and operation flow are integrated to form a failure plan, and the effectiveness and practicability of the plan are ensured through verification, optimization and updating.

[0067] Specifically, based on the failure mode, a set of health state related parameters is defined, a reliable failure criterion is formed, the accuracy of failure diagnosis is improved, and misjudgment and missed judgment are avoided. A clear mapping relationship between the failure mode and the health state parameters is established, so that the system can quickly and accurately judge the failure condition according to the parameter change, and provide strong support for timely handling of failures. According to the failure knowledge and failure criterion, detailed failure countermeasures and operation flow are defined to ensure that the failure can be quickly and effectively handled when it occurs, and the loss caused by the failure is reduced.

[0068] Specifically, the failures in the failure plan can be divided into system-level failures, regular failures, experimental sample safety failures and degradation failures, the unknown failures are out-of-plan failures, and the failure diagnosis types include interface failures, function failures and performance degradation failures.

[0069] Specifically, the in-plan failures can be divided into system-level failures (mainly for severity I and II failures), regular failures, experimental sample safety failures and degradation failures. In addition, there are out-of-plan failures. The in-orbit failures of the application system generally involve system-to-system / system level, experimental cabinet level and single machine level, and the failure diagnosis types generally include interface failures, function failures and performance degradation failures. Function failures, such as those that cannot be located through telemetry analysis, generally also need to be tested through interfaces to accurately locate them. (1) Function failures are generally internal functional abnormalities of the load, such as internal component, board card and component failures; (2) Interface failures generally include power supply, measurement and control, communication interface failures or communication cable failures; (3) Performance degradation failures are generally performance degradation of the load to varying degrees. Health state data includes work parameter data and telemetry data representing the working state of the load, which includes digital and analog types. The work parameters and telemetry data are divided into stable data, sudden change data and gradual change data. Stable data remains basically unchanged or changes within a certain range, such as small fluctuations in the analog voltage value of the load management unit, sudden change data changes significantly over time, such as the charging module of the power supply controller being disturbed by single particles, causing the charging rate and charging current to jump, and gradual change data gradually increases or decreases over time, such as the refrigeration time of the Stirling refrigerator becoming longer, the positioning speed residual error of the GNSS receiver increasing, etc.

[0070] The fault diagnosis module 30 is connected to the data acquisition module 10 and the fault characterization module 20, and is used to analyze the fault mode according to the fault plan and the health status information to obtain abnormal results and alarm, and to diagnose the abnormal results to obtain the fault diagnosis type.

[0071] Specifically, such as Figure 3 As shown, the fault diagnosis module 30 includes:

[0072] Data processing unit 31 is used to extract and analyze features from the health status information to obtain processing results;

[0073] Specifically, the system receives real-time health status information from the fault characterization module, including but not limited to temperature, pressure, vibration frequency, and current intensity. Signal processing techniques (such as wavelet transform and Fourier transform) are used to extract key features, such as temperature change rate, pressure fluctuation amplitude, and vibration frequency variation. Statistical analysis methods (such as mean and standard deviation) are employed to analyze the changing trends and distribution characteristics of these features.

[0074] The dynamic adaptation unit 32 is connected to the data processing unit 31 and is used to dynamically adjust the threshold rules and statistical rules according to the processing results to obtain the adjustment results to adapt to different environments.

[0075] Specifically, adaptive filtering algorithms are used to dynamically adjust thresholds. For example, the Kalman filter algorithm is used to dynamically adjust the fault diagnosis thresholds based on the real-time operating status and historical data of the equipment, adapting to the equipment's operation under different loads and environmental conditions. The impact of environmental factors on equipment operation is considered, and threshold and statistical rules are compensated for. For instance, temperature and humidity have a significant impact on equipment performance; by establishing a mapping relationship between environmental factors and equipment performance, threshold and statistical rules can be compensated for to improve diagnostic accuracy.

[0076] Anomaly diagnosis unit 33, connected to the dynamic adaptation unit 32, is used to generate anomaly result priority sorting based on the processing result and the adjustment result of the fault plan matching analysis, and to use statistical analysis algorithm to perform probability diagnosis on the anomaly result to obtain the fault diagnosis type.

[0077] Specifically, the anomaly diagnosis unit includes:

[0078] The contingency plan matching subunit is used to perform a matching analysis on the fault contingency plan based on the adjustment results and the processing results to generate matching degree data.

[0079] Specifically, the system receives adjusted detection thresholds from the dynamic adaptation unit and feature vectors from the data processing unit. The feature vectors are then matched with preset fault plans to calculate the matching degree. The matching degree represents the similarity between the feature vectors and the fault plan, and is calculated using the following formula:

[0080]

[0081] The matching degree between feature vectors and fault plans is calculated using methods such as cosine similarity and Euclidean distance. Matching degree data is generated to represent the degree of similarity between feature vectors and each fault plan.

[0082] A priority sorting subunit, connected to the pre-plan matching subunit, is used to prioritize the matching results and generate sorting results based on the severity and probability of occurrence of the anomalies.

[0083] Specifically, the matching results are ranked according to the severity of the contingency plans (e.g., catastrophic, severe, moderate, minor). The probability of a fault occurring is assessed by combining the matching degree and historical health status information. The matching results are then comprehensively ranked based on severity and probability of occurrence to generate a priority ranking result. A weighted ranking algorithm is used, with weights determined by severity and probability of occurrence. The priority ranking result is output, representing the priority of each contingency plan.

[0084] The probability diagnosis subunit is connected to the priority sorting subunit to perform probability assessment on the fault diagnosis type based on the sorting result, obtain the probability distribution of each fault diagnosis type, and then determine the fault diagnosis type.

[0085] Specifically, the matching results are comprehensively ranked based on severity and probability of occurrence to generate a priority ranking result. A weighted ranking algorithm is used, with weights determined according to severity and probability of occurrence. The priority ranking result is output, representing the priority of each contingency plan.

[0086] Specifically, feature extraction and analysis accurately identify abnormal features in health status information, improving the accuracy of the fault diagnosis module. Dynamically adjusting detection thresholds allows the system to adapt to different operating environments and conditions, enhancing the robustness of the fault diagnosis module. Prioritizing abnormal results helps managers quickly identify and handle high-risk faults, improving system reliability.

[0087] The experimental control module 40 is used to collect images and videos of experimental samples in real time to obtain health status data, and also to collect environmental parameters and task safety requirements. Based on the environmental parameters and task safety requirements, the health status data is analyzed to obtain experimental control strategies.

[0088] Specifically, cameras are deployed in the experimental area to capture images and videos of the experimental samples in real time, ensuring accurate recording of health status information such as changes in appearance and color. The acquired image and video data undergoes preliminary processing, such as resolution adjustment and format conversion, for subsequent analysis. Simultaneously, the data is stored for future research and traceability. Sensors are used to monitor environmental parameters in the experimental area in real time, such as temperature, humidity, light intensity, and gas composition, ensuring accurate assessment of the experimental environment's impact on the samples. Based on the experimental objectives and requirements, safety requirements during the experiment are analyzed, such as sample stability requirements and the operational safety requirements of the experimental equipment. Computer vision technology is used to analyze the images and videos of the samples to identify changes in their health status. For example, image recognition technology can be used to detect abnormalities such as cracks or discoloration on the sample surface. The results of the sample image and video analysis are combined with environmental parameter monitoring data to comprehensively assess the sample's health status. For example, the impact of temperature changes on sample color changes is analyzed. Based on the comprehensive assessment of the sample's health status and the mission's safety requirements, experimental control strategies are developed. For example, if the sample discolors under high temperature conditions, it may be necessary to adjust the experimental temperature or shorten the sample's exposure time to high temperatures. Based on real-time feedback from the experiment and changes in the health status of the samples, the experimental control strategy is dynamically adjusted to ensure the safety and effectiveness of the experiment.

[0089] Specifically, by monitoring the health status of experimental samples and environmental parameters in real time, the experimental control module can promptly identify potential safety hazards. Based on the health status of the experimental samples and the safety requirements of the mission, it adjusts experimental parameters and operating procedures in real time, optimizing the experimental process. This helps improve the success rate and efficiency of experiments, reducing experimental failures and resource waste caused by unreasonable parameter settings or improper operation.

[0090] The lifespan prediction module 50 is used to analyze the fault contingency plan, the historical health status information and the health status information to predict the remaining lifespan of the equipment and obtain the prediction result.

[0091] Specifically, the lifetime prediction module includes:

[0092] A feature extraction unit is used to extract lifespan-related features from the health status information to obtain extraction results;

[0093] Specifically, this involves analyzing health status information to identify features related to equipment lifespan, such as vibration amplitude of key components, temperature change rate, and current fluctuations. Feature extraction algorithms, such as principal component analysis (PCA) or data mining techniques, are then used to extract key features from the raw data, forming a feature set.

[0094] A degradation modeling unit, connected to the feature extraction unit, is used to perform degradation modeling on key features in the extraction results to obtain a description of the device's health status.

[0095] Specifically, through correlation analysis and expert knowledge, key features that can effectively characterize the degradation state of equipment are selected from the extracted features. These key features include vibration data (such as vibration acceleration, velocity, and displacement), which can be used to detect wear or imbalance in mechanical components; temperature data (such as maximum temperature and rate of temperature change), which can be used to assess the thermal stress of the equipment; current data (such as current intensity and power factor), which can be used to monitor the operating status of motors or other electrical equipment; and chemical data (such as corrosion rate), which is suitable for assessing the impact of chemical reactions or corrosion on equipment lifespan. Linear regression, nonlinear regression, or machine learning methods (such as support vector machines and neural networks) are used to model the changing trends of these key features over time, obtaining a quantitative description of the equipment's health status. The model should be able to reflect the degradation rate and patterns of the equipment.

[0096] The lifespan prediction unit, connected to the feature modeling unit, is used to predict the remaining lifespan based on the fault contingency plan and the health status description to obtain the prediction result.

[0097] Specifically, the lifetime prediction unit includes:

[0098] The threshold determination subunit is used to integrate the failure thresholds of key features in the health status description and fault contingency plan to obtain the integrated result;

[0099] Specifically, a failure threshold refers to the limit value under which equipment or components can function normally under specific conditions. For each critical characteristic, such as vibration amplitude, temperature, and current, the failure threshold may be expressed as the maximum permissible vibration amplitude, the highest operating temperature, the maximum current, etc. These thresholds are typically determined based on technical specifications provided by the equipment manufacturer, industry standards, and historical failure data. For example, the maximum permissible operating temperature of a motor may be set at 120°C; exceeding this temperature is considered a potential risk of overheating failure. Integrating the failure thresholds of critical characteristics in the health status description and fault contingency plans forms a comprehensive threshold system. This may involve weighting the failure thresholds of different characteristics to reflect their varying degrees of impact on the overall health status of the equipment.

[0100] The lifetime prediction subunit is connected to the feature fusion subunit to perform preliminary remaining lifetime prediction based on the integration result to obtain an initial prediction result.

[0101] Specifically, based on the fused feature data, a linear regression model can be used to make a preliminary estimate of the remaining life of the equipment. This assumes an approximate linear relationship between the remaining life and each feature, and the model parameters are estimated using methods such as least squares. Alternatively, nonlinear models, such as multinomial regression or neural networks based on radial basis functions (RBF), can be used to better fit the complex lifespan degradation patterns. A Markov chain model can be used to divide the equipment's health status into multiple discrete states. Based on the transition probability matrix between different states, the time distribution of the equipment's transition from the current state to the failure state (end of life) can be predicted, thus obtaining an estimate of the remaining life. Alternatively, a Bayesian network model can be used, combining prior knowledge of equipment failure and current health status features, to calculate the posterior probability distribution of the equipment under different remaining lifespans through Bayesian inference, achieving a quantitative prediction of the uncertainty of the remaining life.

[0102] The result optimization subunit is connected to the lifetime prediction subunit to perform error analysis and optimization on the initial prediction result to obtain the prediction result.

[0103] Specifically, a Kalman filter algorithm is used to optimize the prediction results. The lifetime prediction process is treated as a dynamic system, where the state equation describing the change in remaining lifetime over time is described by the health degradation law of the equipment, and the observation equation represents the preliminary prediction result. A recursive estimation method is used to correct the prior estimate of the remaining lifetime using the prediction error at the current moment, resulting in a posterior estimate closer to the true value. Particle swarm optimization is then used to optimize the parameters of the prediction model, using the prediction error as a fitness function. A swarm intelligence search algorithm is employed to find the optimal parameter combination that minimizes the prediction error in the parameter space, thereby improving the accuracy of the prediction results.

[0104] Specifically, feature extraction and modeling can accurately capture key features of equipment health status, thereby improving the accuracy of remaining life prediction. This allows for timely detection of potential faults, enabling proactive maintenance or replacement and reducing equipment downtime and repair costs. It also facilitates efficient maintenance planning, optimized spare parts inventory, and lower operating costs. Furthermore, by predicting the remaining life of equipment, proactive measures can be taken to prevent safety incidents caused by equipment failures.

[0105] The scheme update module 60 is connected to the fault diagnosis module 30, the experiment control module 40, and the lifespan prediction module 50. It is used to generate the initial health management scheme based on the fault diagnosis type, the experiment control strategy, and the lifespan prediction result. Based on the health management scheme and the health status information, it uses a machine learning model to perform fault detection on unknown faults to obtain detection results. Based on the detection results, it optimizes the initial health management scheme to obtain the target health management scheme.

[0106] Specifically, the scheme update module includes:

[0107] The scheme integration unit is used to generate an initial health management scheme that includes fault handling strategies, parameter control strategies, and maintenance strategies based on the fault diagnosis type, experimental safety risks, and life prediction results.

[0108] Specifically, the fault diagnosis module obtains fault diagnosis type information, including interface faults, functional faults, and performance degradation faults. The experimental control module obtains experimental safety risk data, including health status assessments of experimental samples (such as image and video analysis results), environmental parameters (such as temperature, humidity, and pressure), and mission safety requirements (such as special requirements for equipment operation). Fault handling strategies: For recoverable faults (such as interface faults), measures are developed to restart the equipment, repair the interface, or isolate the faulty component. For example, for interface faults, a restart procedure is developed: first, disconnect the equipment power, wait 30 seconds, then reconnect the power, and check if the equipment returns to normal operation. For unrecoverable faults (such as performance degradation faults reaching their lifespan limit), a plan is developed to replace the faulty component or equipment. For example, when the performance degradation of a critical component causes the equipment to fail to meet mission requirements and cannot be repaired, the component is scheduled for replacement during the next maintenance window. Parameter adjustment strategies: Based on the equipment's health status and lifespan prediction results, the operating parameters of the equipment are adjusted to optimize performance and extend its service life. For example, for a motor, if the life prediction shows a short remaining lifespan, its load power should be appropriately reduced, operating time shortened, and overload operation avoided to extend the motor's lifespan. Based on experimental safety risk data, parameters during the experiment should be adjusted to ensure experimental safety. For example, if the health status assessment of the experimental sample shows abnormalities, parameters such as temperature and pressure should be appropriately reduced to avoid further damage to the sample. Maintenance and repair strategies: Based on the remaining lifespan and failure risk of the equipment, preventative maintenance and repair plans should be arranged. For example, for equipment X with an expected remaining lifespan of 120 days, preventative maintenance should be arranged 30 days in advance, including checking critical components and replacing vulnerable parts. Maintenance resources should be allocated rationally considering the arrangement of experimental tasks and the frequency of equipment use. For example, more equipment maintenance work should be scheduled during periods of lighter experimental tasks to reduce the impact on experimental tasks. The various strategies formulated above should be integrated to form an initial health management plan that includes fault handling strategies, parameter control strategies, and maintenance and repair strategies. The feasibility and effectiveness of the initial plan should be evaluated to ensure that it meets the health management needs of the equipment and is coordinated with the overall operation plan of the space station.

[0109] A fault detection unit, connected to the scheme integration unit, is used to identify the unknown faults based on the historical health status information using a machine learning model to obtain the detection results.

[0110] Specifically, the fault detection unit includes:

[0111] A model training subunit is connected to the data processing unit to use the processing results as a training set to train the machine learning model to obtain a trained model.

[0112] Specifically, the collected historical health status information is cleaned to remove noisy, erroneous, and duplicate data to improve data quality. The cleaned data is then normalized to convert data of different dimensions and ranges to the same scale for easier subsequent analysis and processing. The preprocessed data is used as a training set, and a suitable machine learning algorithm (such as decision trees, support vector machines, or neural networks) is selected to build a fault detection model. The model is then trained using the training set, and its parameters and structure are adjusted to ensure it can accurately identify known fault modes and possess a certain degree of generalization ability to handle unknown faults.

[0113] The feature engineering subunit is used to analyze the health status information to extract feature vectors for fault detection;

[0114] Specifically, a comprehensive analysis of health status information is conducted to extract feature vectors relevant to fault detection. These features may include equipment operating parameters, performance indicators, and fault symptoms. By comparing and analyzing historical health status information and normal operation data, features with significant distinguishing ability for fault detection are selected, and a set of feature vectors is constructed.

[0115] The result generation subunit is connected to the feature engineering subunit to use the feature vector as input to the training model to obtain the detection result.

[0116] Specifically, the extracted feature vectors are input into the trained machine learning model. The model analyzes and judges based on the input feature vectors and outputs fault detection results, including information such as whether a fault exists, the type of fault, and the severity of the fault.

[0117] The scheme update unit is connected to the fault detection unit and is used to provide feedback on the initial health management scheme based on the detection results to obtain the target health management scheme.

[0118] Specifically, based on the detection results output by the fault detection unit, the shortcomings of the initial health management plan in dealing with unknown faults are analyzed. For example, some fault diagnosis types were not fully considered, or some handling measures were ineffective in practice. Based on the feedback analysis results, the initial health management plan is adjusted and optimized. This may include adding or modifying fault handling strategies, adjusting the parameter range of parameter adjustment strategies, and revising the maintenance plan of facility maintenance strategies. Combining the new fault detection results and the latest health status information of the equipment, the various strategies of the health management plan are re-evaluated and adjusted to ensure that it can better cope with various possible fault situations of the equipment. After multiple iterations of optimization, the target health management plan is finally generated. This plan, based on a full consideration of known and unknown faults, provides more comprehensive and effective health management measures, which can better ensure the reliable operation of the space station's payloads.

[0119] Specifically, by comprehensively considering fault diagnosis types, target health management data, and lifespan prediction results, the generated health management plan can more comprehensively cover various potential equipment failure scenarios, improving the accuracy and effectiveness of fault management. Utilizing machine learning models to detect unknown faults and optimizing the health management plan based on the detection results enables the system to better adapt to uncertainties and changes during equipment operation, improving system flexibility and robustness. Through accurate lifespan prediction and fault detection, facility maintenance and resource allocation can be more rationally scheduled, avoiding unnecessary maintenance and resource waste, and reducing operating costs.

[0120] Specifically, the space-based component transmits the collected health status information to the ground-based component. The ground-based component performs offline training and optimization of the fault characterization and diagnosis model based on the health status information and the historical health status information, and updates the optimized model to the space-based component via the space-ground link. The space-based component performs real-time fault detection, isolation, and recovery of the payload's health status based on the optimized model, and feeds back the execution results to the ground-based component. The ground-based component further adjusts and optimizes the model based on the execution results, realizing dynamic optimization of space-ground collaboration.

[0121] Specifically, the trained model is compressed and converted to reduce its size and computational load, enabling it to run efficiently on the limited computing resources of the space-based component. Common model compression techniques include parameter quantization, pruning, and knowledge distillation. Differences between the latest version of the computational model and the currently running model on the space-based component are incrementally updated by transmitting only the differences via the space-ground link. This method significantly reduces the data transmission volume and time cost required for model updates. After the optimized model is updated to the space-based component via the space-ground link, the fault diagnosis module on the space-based component uses the updated model to perform real-time reasoning and analysis on the collected health status information, quickly detecting potential fault symptoms. The model can output the equipment's health status assessment results and the possible fault types and locations based on the input real-time data. Upon detecting a fault, the solution update module on the space-based component, based on the fault information provided by the model and combined with the equipment's topology and operating principles, quickly locates the fault source and isolates it from other normally operating equipment or systems to prevent further spread of the fault. Depending on the type and severity of the fault, corresponding recovery operations are automatically executed. For example, minor faults can be self-repaired by reconfiguring equipment parameters or restarting relevant modules; for more serious faults, measures such as switching to backup equipment and adjusting system operating mode can be taken according to the preset fault emergency plan to ensure that some key functions of the load system can continue to operate.

[0122] Specifically, this invention, through the collaborative work of a data acquisition module and an experimental control module, comprehensively monitors the health status of equipment and the condition of experimental samples, providing precise data support for fault diagnosis and lifespan prediction. The close cooperation between the fault characterization module and the fault diagnosis module enables rapid and accurate diagnosis and early warning of equipment faults, ensuring reliable equipment operation. The scheme update module, based on fault diagnosis results, experimental control strategies, and lifespan prediction, generates and optimizes health management schemes, enhancing the system's intelligent management level. The close collaboration of these modules achieves space-ground coordinated optimization, significantly improving the accuracy and timeliness of fault diagnosis, enhancing the system's autonomy and reliability, extending equipment lifespan, reducing operating costs, improving resource utilization efficiency, and ensuring the effective operation of space station payloads.

[0123] Specifically, the space-ground integrated autonomous health management system is divided into two parts: space-based and ground-based. Figure 4As shown, the space-based component includes terminal computing power composed of controllers for various experimental cabinets and external payload controllers; system-level computing power composed of information processing systems, portable computers, and intelligent computing units; and application information networks and platform payload networks, forming a heterogeneous computing power platform and network to achieve hierarchical on-orbit autonomous health management. The ground-based component includes a ground control center, a ground mirroring system, and a data center, providing guarantees for algorithm model design, implementation, testing, and verification, including the design, development, and optimization of algorithms and models, and conducting space-to-ground comparisons. The space-based autonomous health management adopts a distributed architecture, distributing health management functions such as fault characterization, fault diagnosis, experimental control, lifespan prediction, and scheme updates to different levels of computing power units. This enables each level of equipment to effectively perform its specific health management tasks, aiming to improve fault response speed during on-orbit operation, reduce the computational burden on the core processor, and enhance the system's autonomy and security. Figure 5 As shown, the distributed architecture comprises three levels: payload level, subsystem level, and system level, encompassing the autonomous health management system from individual payload units to various subsystems, and finally to the entire space station's space application system.

[0124] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0125] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A space station payload earth-space collaborative autonomous health management system, characterized in that, Comprise: a data acquisition module for acquiring real-time telemetry data and operating parameter data of all devices at the load level, subsystem level and system level to obtain health state information; a fault characterization module for obtaining fault knowledge and analyzing to obtain a plurality of fault plans; a fault diagnosis module connected with the data acquisition module and the fault characterization module, for analyzing fault modes according to the fault plans and the health state information to obtain abnormal results and alarm, and diagnosing the abnormal results to obtain a fault diagnosis type; an experimental control module for collecting images and videos of experimental samples in real time to obtain health state data, and collecting environmental parameters and task safety requirements, and analyzing the health state data according to the environmental parameters and task safety requirements to obtain an experimental control strategy; a life prediction module for analyzing the fault plans and the health state information to predict the remaining life of the device to obtain a prediction result; a scheme updating module connected with the fault diagnosis module, the experimental control module and the life prediction module, for generating the initial health management scheme according to the fault diagnosis type, the experimental control strategy and the life prediction result, using a machine learning model to detect unknown faults according to the health management scheme and the health state information to obtain a detection result, and optimizing the initial health management scheme according to the detection result to obtain a target health management scheme.

2. The space station payload ground-space collaborative autonomous health management system according to claim 1, wherein, The faults in the fault plans can be divided into system-level faults, regular faults, experimental sample safety faults and degradation faults, the unknown faults are out-of-plan faults, and the fault diagnosis type includes interface faults, function faults and performance degradation faults.

3. The space station payload ground-space collaborative autonomous health management system according to claim 2, wherein, The fault characterization module comprises: a knowledge acquisition unit for analyzing interface data, fault mode influence analysis and fault tree to obtain fault knowledge; a relationship mapping unit connected with the knowledge acquisition unit, for defining a parameter set related to health state according to the fault knowledge to obtain fault criteria; a plan generation unit connected with the knowledge acquisition unit and the relationship mapping unit respectively, for defining fault response measures and operation processes according to the fault knowledge and the fault criteria, and integrating to form a plurality of fault plans.

4. The space station payload ground-space collaborative autonomous health management system according to claim 3, wherein, The fault diagnosis module comprises: a data processing unit for feature extraction and analysis of the health state information to obtain a processing result; a dynamic adaptation unit connected with the data processing unit, for dynamically adjusting threshold rules and statistical rules according to the processing result to obtain an adjustment result to adapt to different environments; an anomaly diagnosis unit connected with the dynamic adaptation unit, for generating an abnormal result priority order according to the processing result and the adjustment result, and using a statistical analysis algorithm to probabilistically diagnose the abnormal result to obtain the fault diagnosis type.

5. The space station payload ground-space collaborative autonomous health management system according to claim 4, wherein, The anomaly diagnosis unit comprises: a plan matching subunit for matching analysis of fault plans according to the adjustment result and the processing result to generate matching degree data; The priority ranking subunit is connected with the scenario matching subunit, and is configured to perform priority ranking on the matching result, and generate a ranking result based on severity and occurrence probability of the abnormality; The probability diagnosis subunit is connected with the priority ranking subunit, and is configured to perform probability evaluation on the fault diagnosis type according to the ranking result, to obtain a probability distribution of each fault diagnosis type, and to determine the fault diagnosis type.

6. The space station payload ground-space collaborative autonomous health management system according to claim 5, wherein, The life prediction module comprises: The feature extraction unit is configured to extract life-related features from the health state information to obtain an extraction result; The degradation modeling unit is connected with the feature extraction unit, and is configured to perform degradation modeling on key features in the extraction result to obtain a health state description of the equipment; The life prediction unit is connected with the feature modeling unit, and is configured to perform residual life prediction according to the fault scenario and the health state description to obtain the prediction result.

7. The space station payload earth-space collaborative autonomous health management system of claim 6, wherein, The life prediction unit comprises: The threshold determination subunit is configured to integrate failure thresholds of key features in the health state description and the fault scenario to obtain an integration result; The life prediction subunit is connected with the feature fusion subunit, and is configured to perform preliminary residual life prediction according to the integration result to obtain an initial prediction result; The result optimization subunit is connected with the life prediction subunit, and is configured to perform error analysis and optimization on the initial prediction result to obtain the prediction result.

8. The space station payload ground-space collaborative autonomous health management system according to claim 7, wherein, The scheme updating module comprises: The scheme integration unit is configured to generate an initial health management scheme comprising a fault handling strategy, a parameter control strategy and a maintenance and repair strategy according to the fault diagnosis type, the experimental safety risk and the life prediction result; The fault detection unit is connected with the scheme integration unit, and is configured to obtain historical health state information, and to identify the unknown fault using a machine learning model according to the historical health state information to obtain the detection result; The scheme updating unit is connected with the fault detection unit, and is configured to perform feedback on the initial health management scheme according to the detection result to obtain the target health management scheme.

9. The space station payload earth-space collaborative autonomous health management system of claim 8, wherein, The fault detection unit comprises: The model training subunit is configured to perform data preprocessing on the historical health state information, and to use the preprocessed historical health state information as a training set to train the machine learning model to obtain a trained model; The feature engineering subunit is configured to analyze the health state information to extract a feature vector for fault detection; The result generation subunit is connected with the feature engineering subunit, and is configured to use the feature vector as an input of the trained model to obtain the detection result.

10. The space station payload earth-space collaborative autonomous health management system of claim 9, wherein, The space-based part transmits the collected health state information to the ground-based part, the ground-based part performs offline training and optimization of a fault characterization and diagnosis model based on the health state information and the historical health state information, and updates the optimized model to the space-based part through a space-ground link; the space-based part performs real-time fault detection, isolation and recovery on the health state of the payload according to the optimized model, and feeds back an execution result to the ground-based part; The ground-based part further adjusts and optimizes the model according to the execution result, to realize dynamic optimization of space-ground cooperation.

Images