A method and system for spacecraft single-unit fault management based on multi-source data fusion
By employing a multi-source data fusion-based spacecraft fault management method, this approach utilizes an environmental condition decoupling model and a dynamic time-sensitive algorithm to identify latent deviations. Combined with physical information neural networks and thermoelectric calculation optimization, it addresses the communication lag and insufficient multi-dimensional data fusion issues inherent in traditional spacecraft fault management. This enables accurate prediction and autonomous, flexible control, thereby enhancing the spacecraft's autonomous operation and maintenance capabilities.
Patent Information
- Application Number
- CN202611248237.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-18
- Publication Date
- 2026-11-17
AI Technical Summary
Traditional spacecraft fault management methods rely on ground telemetry and control stations, which suffer from large communication delays and slow response times. They are unable to cope with sudden or gradual faults in the space environment and lack the ability to deeply integrate multi-dimensional heterogeneous data, leading to false alarms, missed alarms, and damage to system functions.
A multi-source data fusion method is adopted, and external interference is removed by a two-dimensional decoupling model of environmental conditions. Combined with dynamic time-sensitive weighting algorithm and physical information neural network, implicit deviation identification and remaining life prediction are realized. Elastic degradation control is carried out through thermoelectric-calculation coupled multi-objective optimization.
It improves the accuracy of fault management and autonomous operation and maintenance capabilities of spacecraft in complex space environments, enables accurate identification and reliable prediction of early hidden deviations, and extends the on-orbit service life of the system.
Smart Images

Figure CN123113803A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of spacecraft autonomous operation and maintenance technology, and in particular to a spacecraft single-unit fault management method and system based on multi-source data fusion. Background Technology
[0002] With the increasing complexity of space missions and the continuous extension of on-orbit time, the reliability and autonomous management capabilities of individual spacecraft equipment have become key factors in ensuring mission success.
[0003] Traditional spacecraft fault management methods rely heavily on post-event data analysis and command injection from ground control stations, which suffers from problems such as large communication delays and slow response times, making it difficult to cope with the real-time handling needs of sudden or gradual faults in the space environment. At the same time, existing onboard fault diagnosis technologies often rely on single sensor threshold judgments or simple logical rules, lacking the ability to deeply integrate multi-dimensional heterogeneous data, and are susceptible to external environmental interference such as space radiation, temperature fluctuations, and operating condition switching, resulting in false alarms or missed alarms. In addition, traditional methods usually adopt fixed shutdown or restart strategies after detecting anomalies, lacking a flexible control mechanism to dynamically adjust operating parameters according to the health status of equipment, which means that the system cannot maximize mission performance when some functions are damaged.
[0004] Therefore, there is an urgent need to develop a spacecraft single-unit fault management method and system that can integrate multi-source monitoring data, accurately identify hidden deviations, reliably predict remaining lifespan, and achieve autonomous elastic degradation control, so as to improve the autonomous operation and maintenance level and mission resilience of spacecraft in complex space environments. Summary of the Invention
[0005] In view of the aforementioned existing problems, the present invention is proposed.
[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a spacecraft single-unit fault management method based on multi-source data fusion, comprising: Acquire multi-dimensional raw data of a single spacecraft and perform time alignment and anomaly removal on the multi-dimensional raw data to obtain a multi-source preprocessed data vector; A two-dimensional decoupling model of environmental conditions is constructed and updated using on-board fault-free window data to obtain the updated decoupling matrix. Based on the updated decoupling matrix, interference stripping calculation is performed on multi-source preprocessed data vectors to obtain the intrinsic feature vector of equipment health. The current intrinsic feature sequence is obtained by extracting the sequence of the intrinsic feature vector of the equipment health. A dynamic time sensitivity weight vector is constructed based on the local rate of change distribution of the current intrinsic feature sequence. The dynamic time warping algorithm with the introduction of the dynamic time sensitivity weight vector is used to calculate the weighted optimal bending path distance between the current intrinsic feature sequence and the pre-stored standard health baseline sequence. When the weighted optimal bending path distance is greater than the preset dynamic threshold, a latent deviation trigger signal is generated. A physical information neural network model is constructed in response to a latent deviation trigger signal, and the physical information neural network model is corrected using real-time multi-source monitoring data to obtain a corrected physical information neural network model. Based on the corrected physical information neural network model, the health status decay curve is extrapolated to obtain the predicted value of the remaining service life of the single unit and the slope of the health status decline. The weighted optimal bending path distance, the predicted value of the remaining service life of a single machine, the slope of the decline in health status, and the rate of change of the original key parameters in the multi-source preprocessed data vector are transformed into the first basic probability allocation function, the second basic probability allocation function, and the third basic probability allocation function, respectively. The information entropy and the dynamic trend synergy distribution between the three basic probability allocation functions are calculated. Based on the information entropy and the dynamic trend synergy distribution, a multi-source evidence dynamic conflict penalty matrix is constructed and the comprehensive weight is calculated. The proportional conflict allocation rule is used to perform anti-conflict fusion calculation on the three weighted basic probability allocation functions and the conflict probability is allocated in a targeted manner using the multi-source evidence dynamic conflict penalty matrix to obtain the single machine state confidence probability distribution. When the probability distribution of the confidence level of a single machine meets the preset triggering conditions, a thermoelectric-thermal-computation coupled multi-objective optimization function is constructed with processor frequency, power supply voltage and task load allocation rate as decision variables. The optimal parameter combination is solved and sent to the single machine power management module and task scheduling module to perform elastic degradation control.
[0007] As a preferred embodiment of the spacecraft single-unit fault management method based on multi-source data fusion described in this invention, the specific steps for constructing a two-dimensional decoupling model of environmental conditions and updating the updated decoupling matrix using on-board fault-free window data are as follows: Collect temperature sensor readings, radiation dose rate readings, and single-unit operating current readings during the fault-free window period on the satellite to construct a raw data matrix of environmental conditions. Principal component analysis was performed on the original environmental condition data matrix to extract the principal component vectors of environmental disturbance and operating condition load. Using the principal component vectors of environmental disturbance and operating load as inputs and the multi-source preprocessed data vectors as outputs, a linear regression model is trained to obtain the initial decoupling matrix; By continuously collecting new fault-free data using a sliding time window mechanism, and using incremental least squares method to iteratively correct the initial decoupling matrix online, the updated decoupling matrix is obtained.
[0008] As a preferred embodiment of the spacecraft single-unit fault management method based on multi-source data fusion described in this invention, the specific steps for constructing a dynamic time-sensitivity weight vector based on the local rate of change distribution of the current intrinsic feature sequence are as follows: Calculate the absolute value of the first difference between adjacent data points in the current intrinsic feature sequence to obtain the sequence of absolute values of the first difference; The probability distribution of the rate of change is obtained by normalizing the first-order difference absolute value sequence. The locations of data points in the probability distribution of the rate of change that are higher than a preset rate of change threshold are marked as high-sensitivity intervals; A first weight coefficient is assigned to data points within the high-sensitivity interval, and a second weight coefficient is assigned to data points within the non-high-sensitivity interval. The first weight coefficient is greater than the second weight coefficient. The weight coefficients of all positions are combined to obtain the dynamic time sensitivity weight vector.
[0009] As a preferred embodiment of the spacecraft single-unit fault management method based on multi-source data fusion described in this invention, the specific steps of calculating the weighted optimal curved path distance using a dynamic time warping algorithm incorporating a dynamic time sensitivity weight vector are as follows: Construct the cumulative cost matrix between the current intrinsic feature sequence and the pre-stored standard healthy baseline sequence; During the calculation of each element of the cumulative cost matrix, the weighted cumulative cost is calculated, expressed as follows: ; in, The cumulative cost matrix is represented by the first... Line 1 Column element values, The dynamic time sensitivity weight vector represents the weight vector at the th... The value at each position, Indicates the current intrinsic feature sequence number 1 Data points, Indicates the pre-stored standard healthy baseline sequence number One data point; The dynamic programming algorithm is used to search for the path with the minimum cumulative cost from the starting point to the ending point in the cumulative cost matrix, which is then used as the weighted optimal curved path. The weighted optimal curved path distance is obtained by dividing the sum of all weighted local costs on the weighted optimal curved path by the path length.
[0010] As a preferred embodiment of the spacecraft single-unit fault management method based on multi-source data fusion described in this invention, the specific steps for obtaining the predicted remaining service life and health status decline slope of the single unit by extrapolating the health status decay curve based on the modified physical information neural network model are as follows: By inputting the intrinsic health feature vector of the equipment and real-time multi-source monitoring data into the corrected physical information neural network model, the current health status estimate is obtained. Using the current health status estimate as the initial condition, forward integration is performed using the physical degradation equation embedded in the modified physical information neural network model to generate a health status prediction sequence for future time periods. Linear fitting is performed on the health status prediction sequence to obtain the slope of health status decline; The time difference between the time point when the health status prediction sequence first falls below the preset failure threshold and the current time is calculated as the predicted value of the remaining service life of the single unit.
[0011] As a preferred embodiment of the spacecraft single-unit fault management method based on multi-source data fusion described in this invention, the specific steps of constructing a multi-source evidence dynamic conflict penalty matrix based on information entropy and dynamic trend synergy distribution and calculating the comprehensive weight are as follows: Calculate the information entropy values of the first, second, and third basic probability assignment functions respectively; Calculate the Pearson correlation coefficient between any two basic probability assignment functions as the dynamic trend coherence. Calculate the first The overall weight of each source of evidence: ; in Indicates the first The overall weight of each source of evidence Indicates the first The mean of the dynamic trend correlation between one source of evidence and other sources of evidence. Indicates the first The information entropy value of the basic probability assignment function Represents the conflict penalty factor, when the first When the dynamic trend synergy of individual evidence sources is lower than a preset synergy threshold Take a positive number less than 1, otherwise... Take 1; By using three comprehensive weights to weight and fuse the three corresponding basic probability allocation functions, the single-machine state confidence probability distribution is obtained.
[0012] As a preferred embodiment of the spacecraft single-unit fault management method based on multi-source data fusion described in this invention, the specific steps of constructing a thermoelectric-thermal-computational coupled multi-objective optimization function with processor frequency, power supply voltage, and task load allocation rate as decision variables, solving for the optimal parameter combination, and distributing it to the single-unit power management module and task scheduling module for elastic degradation control are as follows: Establish the thermal resistance network mapping relationship between processor frequency and peak junction temperature, the quadratic polynomial mapping relationship between supply voltage and power consumption, and the performance evaluation mapping relationship between task load allocation rate and task service quality. A Pareto front is constructed with three optimization objectives: minimizing peak junction temperature, minimizing power consumption, and maximizing task service quality. A non-dominated sorting genetic algorithm is used to search for a Pareto optimal solution set within the feasible region defined by the safe temperature constraint and the minimum task service quality constraint. The processor frequency, power supply voltage, and task load allocation rate values corresponding to the solution with the highest task service quality are selected from the Pareto optimal solution set as the optimal parameter combination. The optimal parameter combination is sent to the single-machine power management module and the task scheduling module via the on-board bus. The power management module adjusts the power supply voltage output, and the task scheduling module redistributes the processor frequency and task load allocation rate to complete the elastic degradation control execution.
[0013] Secondly, the present invention provides a spacecraft single-unit fault management system based on multi-source data fusion, comprising: The module includes a data preprocessing module, a feature decoupling module, a deviation detection module, a lifetime prediction module, an evidence fusion module, and a resilience control module. The data preprocessing module is used to acquire multidimensional raw data and perform time alignment and anomaly removal processing to obtain multi-source preprocessed data vectors; The feature decoupling module is used to construct and update the two-dimensional decoupling model of environmental conditions to obtain the updated decoupling matrix, and to remove interference based on the updated decoupling matrix to obtain the intrinsic feature vector of equipment health. The deviation detection module is used to construct a dynamic time-sensitive weight vector and calculate the weighted optimal curved path distance, and generate a hidden deviation trigger signal when the distance exceeds the limit. The lifespan prediction module is used to construct and correct the physical information neural network model, and extrapolate the health status decay curve to obtain the remaining lifespan prediction value and the slope of health status decline. The evidence fusion module is used to transform the basic probability allocation function and calculate the information entropy and dynamic trend synergy, construct the conflict penalty matrix, calculate the comprehensive weight, complete the anti-conflict fusion, and obtain the single-machine state confidence probability distribution. The elastic control module is used to construct a thermoelectric-calculation coupled multi-objective optimization function to solve for the optimal parameter combination, and drive power management and task scheduling to execute elastic degradation control.
[0014] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the spacecraft single-machine fault management method based on multi-source data fusion as described in the first aspect of the present invention.
[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the spacecraft single-unit fault management method based on multi-source data fusion as described in the first aspect of the present invention.
[0016] The beneficial effects of this invention are as follows: by using a two-dimensional decoupling model of environmental conditions to remove external interference and extract the intrinsic health characteristics of the equipment, and combining it with a dynamic time-sensitive weighted regularization algorithm to accurately identify early latent deviations, the invention utilizes the fusion mechanism of physical information neural networks and the advantages of data-driven approaches to achieve reliable prediction of remaining lifetime and degradation trends. Furthermore, based on the synergy between information entropy and dynamic trends, a conflict penalty matrix is constructed to complete the anti-conflict fusion of multi-source evidence to improve the confidence of state assessment. Through thermoelectric and thermal computational coupling multi-objective optimization, elastic degradation control is achieved, thereby improving the accuracy, robustness, and autonomous operation and maintenance capabilities of spacecraft single-unit fault management in complex space environments. Attached Figure Description
[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a flowchart of the spacecraft single-unit fault management method based on multi-source data fusion in Example 1.
[0019] Figure 2 This is a schematic diagram of the spacecraft single-unit fault management system based on multi-source data fusion in Example 1. Detailed Implementation
[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0021] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0022] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.
[0023] Example 1, referring to Figure 1 and Figure 2 This is one embodiment of the present invention, which provides a spacecraft single-unit fault management method based on multi-source data fusion, including the following steps: S1. Acquire multi-dimensional raw data of a single spacecraft and perform time alignment and anomaly removal on the multi-dimensional raw data to obtain a multi-source preprocessed data vector.
[0024] Furthermore, the system collects raw data streams from multi-dimensional sensors of temperature, vibration, current, and voltage on individual spacecraft units, and marks each data point with a high-precision timestamp. Using a preset baseline sampling period as a reference, cubic spline interpolation is employed to resample the raw data streams of each sensor, unifying all channel data to the same time axis to obtain a time-aligned data matrix. The system then calculates the absolute deviation of the median within a sliding window for each channel data in the time-aligned data matrix, and marks data points that deviate from the median by more than a preset multiple of the absolute deviation of the median as outliers. The corresponding outlier is replaced by the linear interpolation result of the adjacent valid data points before and after the outlier to complete the outlier removal process; the data of each channel after outlier removal are concatenated by column and Z-score standardized to obtain a multi-source preprocessed data vector.
[0025] It should be noted that high-precision timestamp marking and cubic spline interpolation resampling have enabled precise alignment of multi-source heterogeneous sensor data on a unified time axis, effectively eliminating the temporal misalignment problem caused by sampling rate differences. At the same time, the anomaly detection and linear interpolation repair mechanism based on median absolute deviation can robustly identify and eliminate transient interference noise. Combined with standardization processing, the data quality and consistency of subsequent feature extraction and model training are ensured.
[0026] S2. Construct a two-dimensional decoupling model of environmental conditions and update the two-dimensional decoupling model of environmental conditions using on-board fault-free window data to obtain the updated decoupling matrix. Based on the updated decoupling matrix, perform interference stripping calculation on the multi-source preprocessed data vector to obtain the intrinsic health feature vector of the equipment.
[0027] Furthermore, temperature sensor readings, radiation dose rate readings, and single-unit operating current readings are collected during the onboard fault-free window to construct an original environmental condition data matrix. Principal component analysis is performed on the original environmental condition data matrix to extract the principal component vectors of environmental disturbance and operating load. Using the principal component vectors of environmental disturbance and operating load as inputs and the multi-source preprocessed data vectors as outputs, a linear regression model is trained to obtain the initial decoupling matrix. New fault-free data are continuously collected using a sliding time window mechanism, and the initial decoupling matrix is iteratively corrected online using the incremental least squares method to obtain the updated decoupling matrix.
[0028] It should be noted that by using principal component analysis to separate the two-dimensional factors of environmental disturbance and operating load, and by using incremental least squares method to iteratively correct the decoupling matrix online, it is possible to adaptively track the coupling relationship drift caused by changes in the external environment and equipment aging during the long-term operation of the spacecraft. This allows for the accurate removal of the influence of non-faulty external disturbances, the extraction of intrinsic features that truly reflect the health status of the equipment, and the improvement of the accuracy and robustness of fault diagnosis.
[0029] S3. Obtain the current intrinsic feature sequence by extracting the sequence of the intrinsic feature vector of the equipment health. Construct a dynamic time sensitivity weight vector based on the local rate of change distribution of the current intrinsic feature sequence. Use a dynamic time warping algorithm that introduces the dynamic time sensitivity weight vector to calculate the weighted optimal bending path distance between the current intrinsic feature sequence and the pre-stored standard health baseline sequence. When the weighted optimal bending path distance is greater than the preset dynamic threshold, generate a latent deviation trigger signal.
[0030] Furthermore, the absolute value of the first difference between adjacent data points in the current intrinsic feature sequence is calculated to obtain the absolute value of the first difference sequence; the absolute value of the first difference sequence is normalized to obtain the probability distribution of the rate of change; and the positions of data points in the probability distribution of the rate of change that are higher than the preset rate of change threshold are marked as high-sensitivity intervals.
[0031] A first weight coefficient is assigned to data points within the high-sensitivity interval, and a second weight coefficient is assigned to data points within the non-high-sensitivity interval. The first weight coefficient is greater than the second weight coefficient. The weight coefficients of all positions are combined to obtain the dynamic time sensitivity weight vector.
[0032] Construct a cumulative cost matrix between the current intrinsic feature sequence and the pre-stored standard healthy baseline sequence; during the calculation of each element of the cumulative cost matrix, calculate the weighted cumulative cost, expressed as: ; in, The cumulative cost matrix is represented by the first... Line 1 Column element values, The dynamic time sensitivity weight vector represents the weight vector at the th... The value at each position, Indicates the current intrinsic feature sequence number 1 Data points, Indicates the pre-stored standard healthy baseline sequence number Data points.
[0033] The dynamic programming algorithm is used to search for the path with the minimum cumulative cost from the starting point to the ending point in the cumulative cost matrix, which is taken as the weighted optimal curved path. The sum of all weighted local costs on the weighted optimal curved path is divided by the path length to obtain the distance of the weighted optimal curved path.
[0034] It should be noted that by dynamically adjusting the time sensitivity weights through the local rate of change distribution, the algorithm pays more attention to the rapidly changing key degradation intervals in the feature sequence, avoiding the insufficient sensitivity problem caused by the traditional dynamic time warping algorithm treating smooth segments and abrupt segments equally. Combined with the weighted optimal curved path distance metric, it can more sensitively capture early hidden deviation trends and achieve timely early warning in the incipient stage of faults.
[0035] S4. In response to the latent deviation trigger signal, a physical information neural network model is constructed and the physical information neural network model is corrected using real-time multi-source monitoring data to obtain the corrected physical information neural network model. Based on the corrected physical information neural network model, the health status decay curve is extrapolated to obtain the predicted value of the remaining service life of the single machine and the slope of the health status decline.
[0036] Furthermore, the intrinsic health feature vector of the equipment and real-time multi-source monitoring data are input into the modified physical information neural network model to obtain the current health status estimate. Using the current health status estimate as the initial condition, forward integration is performed using the physical degradation equation embedded in the modified physical information neural network model to generate a health status prediction sequence for future time periods. The health status prediction sequence is linearly fitted to obtain the health status decline slope. The time difference between the time point when the health status prediction sequence first falls below the preset failure threshold and the current time is calculated as the predicted value of the remaining service life of the single unit.
[0037] It should be noted that by embedding the physical degradation mechanism equation into the neural network architecture and continuously correcting the model parameters through real-time monitoring data, the data-driven method retains its ability to fit complex nonlinear relationships while introducing physical constraints to ensure that the prediction results conform to the actual degradation evolution logic. Thus, even under limited sample conditions, the health state decay curve can still be reliably extrapolated, improving the credibility and engineering applicability of the remaining service life prediction.
[0038] S5. The weighted optimal curved path distance, the predicted value of the remaining service life of a single machine, the slope of the decline in health status, and the rate of change of the original key parameters in the multi-source preprocessed data vector are transformed into the first basic probability allocation function, the second basic probability allocation function, and the third basic probability allocation function, respectively. The information entropy and the dynamic trend synergy distribution between the three basic probability allocation functions are calculated. Based on the information entropy and the dynamic trend synergy distribution, a multi-source evidence dynamic conflict penalty matrix is constructed and the comprehensive weight is calculated. The proportional conflict allocation rule is used to perform anti-conflict fusion calculation on the three weighted basic probability allocation functions and the conflict probability is allocated in a targeted manner using the multi-source evidence dynamic conflict penalty matrix to obtain the single-machine state confidence probability distribution.
[0039] Furthermore, the information entropy values of the first, second, and third basic probability allocation functions are calculated respectively; the Pearson correlation coefficient between any two basic probability allocation functions is calculated as the dynamic trend coherence; and the information entropy values of the third basic probability allocation function are calculated. The overall weight of each source of evidence: ; in Indicates the first The overall weight of each source of evidence Indicates the first The mean of the dynamic trend correlation between one source of evidence and other sources of evidence. Indicates the first The information entropy value of the basic probability assignment function Represents the conflict penalty factor, when the first When the dynamic trend synergy of individual evidence sources is lower than a preset synergy threshold Take a positive number less than 1, otherwise... Set to 1; use three comprehensive weights to perform weighted fusion on the three corresponding basic probability allocation functions to obtain the single-machine state confidence probability distribution.
[0040] It should be noted that by measuring the uncertainty of each evidence source by information entropy and assessing the consistency between evidence by combining dynamic trend synergy, a conflict penalty factor is constructed to adaptively adjust the comprehensive weight of each evidence source. This can effectively suppress the misleading influence of erroneous evidence on the fusion results caused by sensor drift or local failure. Combined with the proportional conflict allocation rule, it can achieve complementary enhancement of multi-source heterogeneous information and improve the confidence of single-machine state assessment and decision reliability.
[0041] S6. When the probability distribution of the confidence level of a single machine meets the preset triggering conditions, a thermoelectric-thermal-calculation coupled multi-objective optimization function is constructed with processor frequency, power supply voltage and task load allocation rate as decision variables. The optimal parameter combination is solved and sent to the single machine power management module and task scheduling module to perform elastic degradation control.
[0042] Furthermore, a thermal resistance network mapping relationship is established between processor frequency and junction peak temperature, a quadratic polynomial mapping relationship between supply voltage and power consumption, and a performance evaluation mapping relationship between task load allocation rate and task service quality. A Pareto front is constructed with the three optimization objectives of minimizing junction peak temperature, minimizing power consumption, and maximizing task service quality. A non-dominated sorting genetic algorithm is used to search for the Pareto optimal solution set within the feasible region defined by the safety temperature constraint and the minimum task service quality constraint.
[0043] The processor frequency, power supply voltage, and task load allocation rate corresponding to the solution with the highest task service quality are selected from the Pareto optimal solution set as the optimal parameter combination. The optimal parameter combination is then sent to the single-machine power management module and the task scheduling module via the on-board bus. This drives the power management module to adjust the power supply voltage output and the task scheduling module to redistribute the processor frequency and task load allocation rate, thus completing the elastic degradation control execution.
[0044] It should be noted that establishing a multi-physics coupling mapping relationship between thermoelectricity and thermal computation and searching for the optimal parameter combination using the Pareto front can collaboratively optimize junction temperature and power consumption while ensuring the baseline of mission service quality. This avoids performance imbalance or thermal safety risks caused by single-objective optimization. Real-time control commands are issued through the onboard bus to achieve linkage response between power management and mission scheduling, giving the spacecraft the ability to autonomously and elastically degrade under abnormal conditions and extending the system's on-orbit usable life.
[0045] This embodiment also provides a spacecraft single-unit fault management system based on multi-source data fusion, including: The system includes a data preprocessing module, a feature decoupling module, a deviation detection module, a lifetime prediction module, an evidence fusion module, and a resilience control module.
[0046] The data preprocessing module is used to acquire multidimensional raw data and perform time alignment and anomaly removal to obtain multi-source preprocessed data vectors.
[0047] The feature decoupling module is used to construct and update the two-dimensional decoupling model of environmental conditions to obtain the updated decoupling matrix, and to obtain the intrinsic feature vector of equipment health by removing interference based on the updated decoupling matrix.
[0048] The deviation detection module is used to construct a dynamic time-sensitive weight vector and calculate the weighted optimal curved path distance, and generate a latent deviation trigger signal when the distance exceeds the limit.
[0049] The lifespan prediction module is used to build and correct the physical information neural network model, and extrapolate the health status decay curve to obtain the remaining lifespan prediction value and the slope of health status decline.
[0050] The evidence fusion module is used to transform the basic probability allocation function and calculate the information entropy and dynamic trend synergy. It constructs a conflict penalty matrix, calculates the comprehensive weight, and completes anti-conflict fusion to obtain the single-machine state confidence probability distribution.
[0051] The elastic control module is used to construct a thermoelectric-calculation coupled multi-objective optimization function to solve for the optimal parameter combination, and drive power management and task scheduling to execute elastic degradation control.
[0052] This embodiment also provides a computer device applicable to the spacecraft single-unit fault management method based on multi-source data fusion, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the spacecraft single-unit fault management method based on multi-source data fusion as proposed in the above embodiment.
[0053] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device's casing, or an external keyboard, touchpad, or mouse.
[0054] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the spacecraft single-unit fault management method based on multi-source data fusion as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.
[0055] In summary, this invention extracts the intrinsic health characteristics of equipment by stripping away external interference through a two-dimensional decoupling model of environmental conditions, accurately identifies early latent deviations by combining a dynamic time-sensitive weighted regularization algorithm, reliably predicts remaining lifetime and degradation trends by utilizing the fusion mechanism of physical information neural networks and the advantages of data-driven approaches, constructs a conflict penalty matrix based on the synergy of information entropy and dynamic trends to complete multi-source evidence anti-conflict fusion to improve the confidence of state assessment, and achieves elastic degradation control through multi-objective optimization of thermoelectric and thermal computational coupling. Thus, it improves the accuracy, robustness, and autonomous operation and maintenance capabilities of spacecraft single-unit fault management in complex space environments.
[0056] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A spacecraft single-unit fault management method based on multi-source data fusion, characterized in that: include: Acquire multi-dimensional raw data of a single spacecraft and perform time alignment and anomaly removal on the multi-dimensional raw data to obtain a multi-source preprocessed data vector; A two-dimensional decoupling model of environmental conditions is constructed and updated using on-board fault-free window data to obtain the updated decoupling matrix. Based on the updated decoupling matrix, interference stripping calculation is performed on multi-source preprocessed data vectors to obtain the intrinsic feature vector of equipment health. The current intrinsic feature sequence is obtained by extracting the sequence of the intrinsic feature vector of the equipment health. A dynamic time sensitivity weight vector is constructed based on the local rate of change distribution of the current intrinsic feature sequence. The dynamic time warping algorithm with the introduction of the dynamic time sensitivity weight vector is used to calculate the weighted optimal bending path distance between the current intrinsic feature sequence and the pre-stored standard health baseline sequence. When the weighted optimal bending path distance is greater than the preset dynamic threshold, a latent deviation trigger signal is generated. A physical information neural network model is constructed in response to a latent deviation trigger signal, and the physical information neural network model is corrected using real-time multi-source monitoring data to obtain a corrected physical information neural network model. Based on the corrected physical information neural network model, the health status decay curve is extrapolated to obtain the predicted value of the remaining service life of the single unit and the slope of the health status decline. The weighted optimal bending path distance, the predicted value of the remaining service life of a single machine, the slope of the decline in health status, and the rate of change of the original key parameters in the multi-source preprocessed data vector are transformed into the first basic probability allocation function, the second basic probability allocation function, and the third basic probability allocation function, respectively. The information entropy and the dynamic trend synergy distribution between the three basic probability allocation functions are calculated. Based on the information entropy and the dynamic trend synergy distribution, a multi-source evidence dynamic conflict penalty matrix is constructed and the comprehensive weight is calculated. The proportional conflict allocation rule is used to perform anti-conflict fusion calculation on the three weighted basic probability allocation functions and the conflict probability is allocated in a targeted manner using the multi-source evidence dynamic conflict penalty matrix to obtain the single machine state confidence probability distribution. When the probability distribution of the confidence level of a single machine meets the preset triggering conditions, a thermoelectric-thermal-computation coupled multi-objective optimization function is constructed with processor frequency, power supply voltage and task load allocation rate as decision variables. The optimal parameter combination is solved and sent to the single machine power management module and task scheduling module to perform elastic degradation control.
2. The spacecraft single-unit fault management method based on multi-source data fusion as described in claim 1, characterized in that: The specific steps for constructing a two-dimensional decoupling model of environmental conditions and updating the updated decoupling matrix using on-board fault-free window data are as follows: Collect temperature sensor readings, radiation dose rate readings, and single-unit operating current readings during the fault-free window period on the satellite to construct a raw data matrix of environmental conditions. Principal component analysis was performed on the original environmental condition data matrix to extract the principal component vectors of environmental disturbance and operating condition load. Using the principal component vectors of environmental disturbance and operating load as inputs, and the multi-source preprocessed data vectors as outputs, a linear regression model is trained to obtain the initial decoupling matrix; By continuously collecting new fault-free data using a sliding time window mechanism, and using incremental least squares method to iteratively correct the initial decoupling matrix online, the updated decoupling matrix is obtained.
3. The spacecraft single-unit fault management method based on multi-source data fusion as described in claim 1, characterized in that: The specific steps for constructing the dynamic time-sensitivity weight vector based on the local rate of change distribution of the current intrinsic feature sequence are as follows: Calculate the absolute value of the first difference between adjacent data points in the current intrinsic feature sequence to obtain the sequence of absolute values of the first difference; The probability distribution of the rate of change is obtained by normalizing the first-order difference absolute value sequence. The locations of data points in the probability distribution of the rate of change that are higher than a preset rate of change threshold are marked as high-sensitivity intervals; A first weight coefficient is assigned to data points within the high-sensitivity interval, and a second weight coefficient is assigned to data points within the non-high-sensitivity interval. The first weight coefficient is greater than the second weight coefficient. The weight coefficients of all positions are combined to obtain the dynamic time sensitivity weight vector.
4. The spacecraft single-unit fault management method based on multi-source data fusion as described in claim 1, characterized in that: The dynamic time warping algorithm, which incorporates a dynamic time-sensitive weight vector, is used to calculate the weighted optimal curved path distance. The specific steps are as follows: Construct the cumulative cost matrix between the current intrinsic feature sequence and the pre-stored standard healthy baseline sequence; During the calculation of each element of the cumulative cost matrix, the weighted cumulative cost is calculated, expressed as follows: ; in, The cumulative cost matrix is represented by the first... Line 1 Column element values, The dynamic time sensitivity weight vector represents the weight vector at the th... The value at each position, Indicates the current intrinsic feature sequence number 1 Data points, Indicates the pre-stored standard healthy baseline sequence number. One data point; The dynamic programming algorithm is used to search for the path with the minimum cumulative cost from the starting point to the ending point in the cumulative cost matrix, which is then used as the weighted optimal curved path. The weighted optimal curved path distance is obtained by dividing the sum of all weighted local costs on the weighted optimal curved path by the path length.
5. The spacecraft single-unit fault management method based on multi-source data fusion as described in claim 1, characterized in that: The method of extrapolating the health status decay curve based on the modified physical information neural network model to obtain the predicted value of the remaining service life of a single machine and the slope of the health status decline is as follows: By inputting the intrinsic health feature vector of the equipment and real-time multi-source monitoring data into the corrected physical information neural network model, the current health status estimate is obtained. Using the current health status estimate as the initial condition, forward integration is performed using the physical degradation equation embedded in the modified physical information neural network model to generate a health status prediction sequence for future time periods. Linear fitting is performed on the health status prediction sequence to obtain the slope of health status decline; The time difference between the time point when the health status prediction sequence first falls below the preset failure threshold and the current time is calculated as the predicted value of the remaining service life of the single unit.
6. The spacecraft single-unit fault management method based on multi-source data fusion as described in claim 1, characterized in that: The specific steps for constructing a dynamic conflict penalty matrix for multi-source evidence based on information entropy and dynamic trend synergy distribution, and calculating the comprehensive weight, are as follows: Calculate the information entropy values of the first, second, and third basic probability assignment functions respectively; Calculate the Pearson correlation coefficient between any two basic probability assignment functions as the dynamic trend coherence. Calculate the first The overall weight of each source of evidence: ; in Indicates the first The overall weight of each source of evidence Indicates the first The mean of the dynamic trend correlation between one source of evidence and other sources of evidence. Indicates the first The information entropy value of the basic probability assignment function Represents the conflict penalty factor, when the first When the dynamic trend synergy of individual evidence sources is lower than a preset synergy threshold Take a positive number less than 1, otherwise... Take 1; By using three comprehensive weights to weight and fuse the three corresponding basic probability allocation functions, the single-machine state confidence probability distribution is obtained.
7. The spacecraft single-unit fault management method based on multi-source data fusion as described in claim 1, characterized in that: The process involves constructing a thermoelectric-thermal-computational coupled multi-objective optimization function with processor frequency, power supply voltage, and task load allocation rate as decision variables. The optimal parameter combination is then obtained and distributed to the single-machine power management module and task scheduling module for elastic degradation control. The specific steps are as follows: Establish the thermal resistance network mapping relationship between processor frequency and peak junction temperature, the quadratic polynomial mapping relationship between supply voltage and power consumption, and the performance evaluation mapping relationship between task load allocation rate and task service quality. A Pareto front is constructed with the three optimization objectives of minimizing junction temperature peak, minimizing power consumption, and maximizing task service quality. A non-dominated sorting genetic algorithm is used to search for a Pareto optimal solution set within the feasible region defined by the safe temperature constraint and the minimum task service quality constraint. The processor frequency, power supply voltage, and task load allocation rate values corresponding to the solution with the highest task service quality are selected from the Pareto optimal solution set as the optimal parameter combination. The optimal parameter combination is sent to the single-machine power management module and the task scheduling module via the on-board bus. The power management module adjusts the power supply voltage output, and the task scheduling module redistributes the processor frequency and task load allocation rate to complete the elastic degradation control execution.
8. A spacecraft single-unit fault management system based on multi-source data fusion, employing the spacecraft single-unit fault management method based on multi-source data fusion as described in any one of claims 1 to 7, characterized in that: include: The module includes a data preprocessing module, a feature decoupling module, a deviation detection module, a lifetime prediction module, an evidence fusion module, and a resilience control module. The data preprocessing module is used to acquire multidimensional raw data and perform time alignment and anomaly removal processing to obtain multi-source preprocessed data vectors; The feature decoupling module is used to construct and update the two-dimensional decoupling model of environmental conditions to obtain the updated decoupling matrix, and to remove interference based on the updated decoupling matrix to obtain the intrinsic feature vector of equipment health. The deviation detection module is used to construct a dynamic time-sensitive weight vector and calculate the weighted optimal curved path distance, and generate a hidden deviation trigger signal when the distance exceeds the limit. The lifespan prediction module is used to construct and correct the physical information neural network model, and extrapolate the health status decay curve to obtain the remaining lifespan prediction value and the slope of health status decline. The evidence fusion module is used to transform the basic probability allocation function and calculate the information entropy and dynamic trend synergy, construct the conflict penalty matrix, calculate the comprehensive weight, complete the anti-conflict fusion, and obtain the single-machine state confidence probability distribution. The elastic control module is used to construct a thermoelectric-calculation coupled multi-objective optimization function to solve for the optimal parameter combination, and drive power management and task scheduling to execute elastic degradation control.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the spacecraft single-machine fault management method based on multi-source data fusion as described in any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the spacecraft single-machine fault management method based on multi-source data fusion as described in any one of claims 1 to 7.
