A method for fault diagnosis of a reaction flywheel

Abstract

The application provides a fault diagnosis method for a reaction flywheel, comprising the following steps: collecting operation data of the reaction flywheel, and pre-processing the operation data; obtaining possible faults and fault occurrence probabilities of the reaction flywheel according to the pre-processed operation data; constructing a fault diagnosis path of the reaction flywheel; processing the fault diagnosis path based on the possible faults and the fault occurrence probabilities by using a grey wolf optimization algorithm to obtain an order of the fault diagnosis path; and performing fault diagnosis based on the order result; after components of the reaction flywheel used for fault diagnosis are randomly combined to form the fault diagnosis path, the fault diagnosis path is compared and sorted by using the grey wolf optimization algorithm, and when fault diagnosis is performed according to the sorting result, the fault diagnosis efficiency can be greatly improved, and the fault position can be quickly located, so that timely maintenance and repair can be performed, and the normal operation of a spacecraft is avoided from being affected.

Description

Technical Field

[0001] This invention relates to the field of aerospace technology, and in particular to a fault diagnosis method for reaction flywheels. Background Technology

[0002] In the high-precision field of aerospace technology, the reaction flywheel is undoubtedly a crucial component for satellite attitude control and orbit adjustment. By cleverly adjusting its own angular momentum, it precisely generates reaction torque, thereby achieving fine adjustment of satellite attitude and flexible maneuvering of orbit. It is the core guarantee for the stable operation of satellites in space and the successful execution of various missions. Its stability and reliability directly determine the ultimate success or failure of satellite missions.

[0003] However, the space environment is full of complex and unpredictable harsh factors. Operating in such an extreme environment for a long time, the reaction flywheel will inevitably suffer from problems such as wear and aging. During its operation, various malfunctions will also occur from time to time. For example, communication anomalies will cause the information transmission between the ground control center and the flywheel to be blocked, making it impossible to receive and send critical commands in a timely manner. There is also the dangerous situation of uncontrolled rotation speed, which may cause the satellite's attitude to become unbalanced and chaotic in an instant. If these malfunctions are not detected and handled properly in time, the operational safety of the satellite will be greatly threatened, and the execution of its mission will be out of the question. All efforts may be in vain.

[0004] Traditional methods for diagnosing reaction flywheel faults primarily rely on ground-based telemetry and remote control technology, as well as manual analysis. Ground control personnel receive limited data transmitted from satellites and rely on their experience and professional knowledge to analyze and make judgments. However, this method has significant drawbacks. Its response speed is often slow, and it cannot accurately reflect the actual operating status of the flywheel in real time. Especially in highly challenging mission scenarios such as deep space exploration, traditional fault diagnosis methods are inadequate due to communication delays caused by long distances and limitations in data transmission. They are unable to meet mission requirements and are gradually becoming inadequate in the rapidly developing wave of aerospace technology.

[0005] In recent years, the rapid rise of artificial intelligence technology has brought new hope to the fault diagnosis of reaction flywheels. By installing various advanced sensors on the reaction flywheel, massive amounts of real-time data can be collected. Through in-depth mining and analysis by artificial intelligence algorithms, it is possible to preliminarily confirm whether there are potential faults in the reaction flywheel. However, due to the numerous components and complex structure of the reaction flywheel itself, fault diagnosis work can mostly be carried out by random exploration, and a clear and orderly diagnostic path has not yet been formed. This has led to a significant increase in the diagnosis time, forcing the reaction flywheel to operate under fault conditions for a long time. As a result, its service life will inevitably be greatly reduced. At the same time, it will also have a serious negative impact on the normal operation of the entire spacecraft, which may trigger a series of chain reactions and threaten the safety and smooth progress of the entire space mission. Summary of the Invention

[0006] Therefore, this invention proposes a fault diagnosis method for reaction flywheels, which can select fault diagnosis paths to improve the efficiency of fault diagnosis.

[0007] The technical solution of this invention is implemented as follows:

[0008] A fault diagnosis method for a reaction flywheel includes the following steps:

[0009] Step S1: Collect the operating data of the reaction flywheel and preprocess the operating data;

[0010] Step S2: Based on the preprocessed operating data, obtain the possible faults of the reaction flywheel and the probability of fault occurrence;

[0011] Step S3: Construct the fault diagnosis path for the reaction flywheel;

[0012] Step S4: The gray wolf optimization algorithm is used to process the fault diagnosis path based on the possible faults and the probability of fault occurrence, and the fault diagnosis path is sorted.

[0013] Step S5: Perform fault diagnosis based on the sorting results.

[0014] Preferably, step S1 includes the following steps:

[0015] Step S11: Collect real-time speed, motor current, bearing temperature and vibration data of the reaction flywheel;

[0016] Step S12: Perform data cleaning, noise reduction, and feature extraction on the collected real-time speed, motor current, bearing temperature, and vibration data.

[0017] Preferably, step S2 includes the following specific steps:

[0018] Step S21: Construct a neural network model and train it to map the normal range of the reaction flywheel operation data into the neural network model;

[0019] Step S22: Input the running data into the neural network model, and let the neural network model process the running data;

[0020] Step S23: The neural network model outputs the possible faults of the reaction flywheel and the probability of the fault occurring.

[0021] Preferably, the specific steps for constructing and training the neural network model in step S21 are as follows: acquiring historical data of the reaction flywheel, which includes historical operating data and fault records of the reaction flywheel; dividing the historical operating data and fault records into a set of data and dividing them into a training set and a test set in a 7:3 ratio; training the neural network using the training set; and testing the neural network using the test set after training; and obtaining the trained neural network model after the test accuracy reaches a set threshold.

[0022] Preferably, the specific steps of step S3 are as follows:

[0023] Step S31: Obtain several diagnostic components for fault diagnosis of the reaction flywheel;

[0024] Step S32: Construct a virtual sphere and uniformly map the diagnostic components onto the surface of the virtual sphere;

[0025] Step S33: Starting from the center of the virtual sphere, randomly select at least one diagnostic component on the surface of the virtual sphere and connect them sequentially to form several fault diagnosis paths.

[0026] Preferably, step S3 further includes:

[0027] Step S34: Based on the possible faults of the reaction flywheel, remove irrelevant parts from the fault diagnosis path.

[0028] Preferably, step S4 includes the following specific steps:

[0029] Step S41: Select potential faults based on the probability of occurrence and the severity of the fault;

[0030] Step S42: Initialize the gray wolf population and set the number of iterations;

[0031] Step S43: Randomly select a fault diagnosis path and calculate the fitness value based on the possible faults;

[0032] Step S44: Select another fault diagnosis path, calculate the fitness value, compare it with the previous calculation result, and sort it according to the comparison result to obtain the sorting result.

[0033] Preferably, step S41 includes the following specific steps:

[0034] Step S411: Assess and quantify the severity of the potential impact of the failure;

[0035] Step S412: Assign values ​​to the probability of failure and the quantified degree of impact, and sum them up.

[0036] Step S413: Select the possible faults from largest to smallest based on the summation result.

[0037] Preferably, the number of iterations in step S42 is the number of fault diagnosis paths.

[0038] Preferably, the specific steps of step S5 are as follows: select a fault diagnosis path from largest to smallest according to the sorting results and perform fault diagnosis.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] 1. This invention provides a fault diagnosis method for reaction flywheels. After collecting the operating data of the reaction flywheel, the possible faults and their probability of occurrence can be evaluated based on the operating data. Then, a complete fault diagnosis path is constructed, and the Grey Wolf optimization algorithm is used to process the fault diagnosis path. The processing is based on the possible faults of the reaction flywheel, thereby obtaining a ranking result of multiple fault diagnosis paths for diagnosing possible faults. The fault diagnosis path with the largest ranking result is the path most likely to diagnose the fault, which can greatly reduce the fault diagnosis time and improve the fault diagnosis efficiency, so as to quickly locate the fault point of the reaction flywheel and carry out timely maintenance to ensure the normal operation of the spacecraft.

[0041] 2. When constructing the fault diagnosis path for the reaction flywheel, a virtual sphere is first constructed. Then, several diagnostic components used for fault diagnosis are mapped onto the surface of the virtual sphere. Starting from the center of the virtual sphere, at least one diagnostic component on its surface is connected in sequence to form several fault diagnosis paths. The virtual sphere ensures that the distance to the first diagnostic component after starting from the center is consistent, and the formed fault diagnosis paths can be displayed intuitively, facilitating subsequent querying of the constructed fault diagnosis paths for data analysis. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in 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 preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart of a fault diagnosis method for a reaction flywheel according to the present invention;

[0044] Figure 2 This is a flowchart of step S1 of a fault diagnosis method for a reaction flywheel according to the present invention;

[0045] Figure 3 This is a flowchart of step S2 of a fault diagnosis method for a reaction flywheel according to the present invention;

[0046] Figure 4 This is a flowchart of step S3 in a fault diagnosis method for a reaction flywheel according to the present invention;

[0047] Figure 5 This is a flowchart of step S4 in a fault diagnosis method for a reaction flywheel according to the present invention;

[0048] Figure 6 This is a flowchart of step S41 of a fault diagnosis method for a reaction flywheel according to the present invention;

[0049] Figure 7 This is a schematic diagram of a virtual sphere used in a fault diagnosis method for a reaction flywheel according to the present invention. Detailed Implementation

[0050] To better understand the technical content of this invention, a specific embodiment is provided below, and the invention will be further described in conjunction with the accompanying drawings.

[0051] See Figures 1 to 7 The present invention provides a fault diagnosis method for a reaction flywheel, comprising the following steps:

[0052] Step S1: Collect the operating data of the reaction flywheel and preprocess the operating data;

[0053] The specific steps include:

[0054] Step S11: Collect real-time speed, motor current, bearing temperature and vibration data of the reaction flywheel;

[0055] Step S12: Perform data cleaning, noise reduction, and feature extraction on the collected real-time speed, motor current, bearing temperature, and vibration data.

[0056] The reaction flywheel is a crucial component of spacecraft, used for adjusting and controlling the spacecraft's attitude and orbit. Therefore, its proper operation is paramount. Installing several sensor arrays on the reaction flywheel allows for the collection of daily operational data. This data enables rapid assessment of any malfunctions. Key data collected include the flywheel's real-time rotational speed, motor current, bearing temperature, and vibration. Abnormal rotational speed may indicate bearing seizure, coupling wear, motor failure, or power system malfunction. Abnormal motor current may suggest motor failure, drive circuit failure, or control system failure. Abnormal bearing temperature may indicate bearing wear, poor lubrication, seizure, or motor failure. Abnormal vibration data may indicate flywheel malfunction, loose mounting base, motor failure, or bearing failure. By collecting the reaction flywheel's operational data, a general assessment of any malfunctions can be made.

[0057] To ensure the operational data can be processed by subsequent neural network models, this invention preprocesses the operational data, including data cleaning, noise reduction, and feature extraction. Data cleaning includes: missing value handling: checking for missing values ​​in the data and selecting appropriate processing methods based on the dataset's attributes and the location of missing values, such as deletion, interpolation, or model imputation; outlier detection and handling: detecting outliers using visualization methods such as histograms and box plots, and selecting appropriate processing methods based on business needs, such as deletion or correction; duplicate value detection and handling: detecting duplicate values ​​by comparing unique values ​​and total counts in the dataset, and using processing methods such as deletion, merging, or modification; and data format standardization: ensuring consistent data formats in the dataset, such as date formats and numeric types, to facilitate subsequent data analysis and processing. Data cleaning improves the accuracy and reliability of the data, providing a reliable foundation for subsequent data analysis, while removing redundant and erroneous data, simplifying the data processing process, and improving efficiency.

[0058] Denoising includes at least one of mean filtering, median filtering, Gaussian filtering, and adaptive filtering. It can remove irrelevant noise from the data, reduce the interference of noise on the data analysis results, and make the data smoother, which is beneficial for subsequent data analysis and modeling.

[0059] Feature extraction can extract the most representative information from raw data, providing strong support for subsequent data analysis and modeling, reducing data dimensionality, decreasing redundant information, improving data processing efficiency and model performance, so that neural network models can process data quickly.

[0060] Step S2: Based on the preprocessed operating data, obtain the possible faults of the reaction flywheel and the probability of fault occurrence;

[0061] The specific steps include:

[0062] Step S21: Construct a neural network model and train it to map the normal range of the reaction flywheel operation data into the neural network model;

[0063] Step S22: Input the running data into the neural network model, and let the neural network model process the running data;

[0064] Step S23: The neural network model outputs the possible faults of the reaction flywheel and the probability of the fault occurring.

[0065] After obtaining the operational data, this invention employs a deep learning model to process the data. This deep learning model utilizes a neural network model, a complex network system formed by the extensive interconnection of numerous simple processing units (called neurons). It is a highly complex nonlinear dynamic learning system. These neurons transmit information to each other through weighted connections and can adjust their connection weights based on the input information and the current state of the network, thereby achieving learning and adaptation. Neural networks possess large-scale parallelism, distributed storage and processing, self-organization, self-adaptation, and self-learning capabilities, making them particularly suitable for handling imprecise and fuzzy information processing problems that require simultaneous consideration of many factors and conditions. A trained neural network model can be obtained through pre-training. Then, the normal values ​​of the corresponding operational data are mapped into the neural network model, which can then process the operational data. The operational data serves as the input to the neural network model, and the output of the model represents potential faults and the probability of each potential fault occurring. This invention introduces a neural network that can identify characteristic patterns of reaction flywheel faults by learning from a large amount of fault data. Even in complex spatial environments, it can quickly and accurately detect reaction flywheel faults. Furthermore, through learning and training, the neural network model can automatically extract fault features, reducing the complexity of fault diagnosis. This makes the fault diagnosis process simpler, more efficient, and easier to implement.

[0066] Preferably, the specific steps for constructing and training the neural network model in step S21 are as follows: acquiring historical data of the reaction flywheel, which includes historical operating data and fault records of the reaction flywheel; dividing the historical operating data and fault records into a set of data and dividing them into a training set and a test set in a 7:3 ratio; training the neural network using the training set; and testing the neural network using the test set after training; and obtaining the trained neural network model after the test accuracy reaches a set threshold.

[0067] A systematic review of the rich historical data accumulated by the reaction flywheel was conducted, and the data was rigorously and meticulously divided according to a specific fixed ratio to form two key datasets: a training set and a test set. Both sets comprehensively cover the massive amount of operational data generated by the reaction flywheel during its long history of operation. This data records in detail the changes of various parameters of the flywheel under different working conditions and different operating stages. In addition, it also includes corresponding fault records, which clearly indicate the specific circumstances and relevant data characteristics of the faults that occurred in the past.

[0068] Once the neural network model is built, its core training phase begins, and it delves into in-depth learning of the training set data. During this process, the model undergoes repeated training based on a complex algorithm architecture and the data in the training set. Each training iteration is a fine-tuning and optimization of the model's internal parameters, continuously improving the model's understanding and grasp of data features. This enables it to gradually and accurately capture the potential correlations and patterns between running data and faults. As training continues, the performance of the neural network model also gradually improves and evolves.

[0069] Once the training process reaches a certain stage, accuracy testing can be conducted using a test set. If the test accuracy of the neural network model on the test set can successfully reach the pre-set threshold standard, it can be officially put into use. The neural network model can receive various data during the operation of the reaction flywheel in real time and quickly and accurately determine whether there are potential faults, providing strong technical support and guarantee for ensuring the stable operation of the reaction flywheel and even the entire satellite system. In addition, during the operation of the reaction flywheel, real-time data can be acquired to continuously update the neural network model and gradually improve its fault diagnosis accuracy.

[0070] Step S3: Construct the fault diagnosis path for the reaction flywheel;

[0071] The specific steps are as follows:

[0072] Step S31: Obtain several diagnostic components for fault diagnosis of the reaction flywheel;

[0073] Step S32: Construct a virtual sphere and uniformly map the diagnostic components onto the surface of the virtual sphere;

[0074] Step S33: Starting from the center of the virtual sphere, randomly select at least one diagnostic component on the surface of the virtual sphere and connect them sequentially to form several fault diagnosis paths.

[0075] Step S34: Based on the possible faults of the reaction flywheel, remove irrelevant parts from the fault diagnosis path.

[0076] After confirming the potential faults and their probabilities in the reaction flywheel, it is necessary to determine which components of the reaction flywheel to start fault diagnosis from. Since some components are interconnected, for certain faults, it is necessary to confirm that the preceding component is functioning correctly before proceeding to the next component to determine the fault location. Therefore, it is necessary to confirm fault diagnosis paths for different components. In constructing these fault diagnosis paths, this invention uses a spherical construction method. First, a virtual sphere is built, with its center as the initial node. After obtaining several diagnostic components for fault diagnosis of the reaction flywheel, these components are evenly distributed across the surface of the virtual sphere as other nodes. Starting from the initial node, one can reach any node containing a diagnostic component. At least two nodes can form a fault diagnosis path. Then, from the node containing the first diagnostic component, one can reach other nodes, ultimately forming several fault diagnosis paths. Each fault diagnosis path contains at least two nodes. However, some noise, i.e., paths completely unrelated to potential faults, may be present in the selected fault diagnosis paths. Removing these paths reduces the number of fault diagnosis paths and lowers the subsequent computational load.

[0077] The distance from the center of the virtual sphere to the surface of the sphere is the same for all paths. This means that the distance from the initial node to the first node in all fault diagnosis paths is also the same. Since the initial node does not need to perform fault diagnosis, the time between the initial node and the first node is set to be uniform by setting the virtual sphere, so that the Gray Wolf optimization algorithm can make a uniform comparison in the future.

[0078] Step S4: The gray wolf optimization algorithm is used to process the fault diagnosis path based on the possible faults and the probability of fault occurrence, and the fault diagnosis path is sorted.

[0079] The specific steps of S4 include:

[0080] Step S41: Select potential faults based on the probability of occurrence and the severity of the fault;

[0081] The specific steps include:

[0082] Step S411: Assess and quantify the severity of the potential impact of the failure;

[0083] Step S412: Assign values ​​to the probability of failure and the quantified degree of impact, and sum them up.

[0084] Step S413: Select the possible faults from largest to smallest based on the summation result.

[0085] Before the Grey Wolf optimization algorithm sorts the fault diagnosis paths, it is first necessary to determine the possible faults to be selected. Since there are multiple possible faults, each of which will be dealt with according to its corresponding severity of impact, the selection order of possible faults is crucial. After obtaining the probability of occurrence of each possible fault through the neural network model, this invention will also evaluate the severity of the impact of possible faults. Then, the importance of possible faults will be evaluated by combining the severity and the probability of occurrence. Faults with a high probability of occurrence and a severe impact need to be dealt with first. If the selection order is inappropriate, it may lead to a lot of effort being spent on some relatively minor fault investigation and diagnosis under limited resources, thus delaying the timely treatment of critical faults and causing serious consequences.

[0086] When assessing the severity of potential fault impacts, an empirical analysis method can be used, combining historical data to evaluate the severity. This is then quantified through expert scoring to obtain the quantified severity. Since the numerical ranges, measurement standards, and data properties involved in fault occurrence probability and severity are often inconsistent, a specially designed assignment rule is needed to convert them into a comparable numerical system. For example, the fault occurrence probability can be assigned a value based on its frequency of occurrence in historical data, the fluctuation range under the influence of relevant environmental factors, and other factors. Similarly, the quantified severity is assigned a value based on its importance weight in the overall system architecture and its interference with critical task execution. After assignment, these two sets of data are summed. This summation yields a comprehensive numerical index reflecting the importance of the fault, i.e., the fault's importance level. When selecting potential faults for processing, the Gray Wolf optimization algorithm selects them in descending order of importance to facilitate fault diagnosis paths for faults with higher importance.

[0087] Step S42: Initialize the gray wolf population and set the number of iterations, which is equal to the number of fault diagnosis paths;

[0088] Step S43: Randomly select a fault diagnosis path and calculate the fitness value based on the possible faults;

[0089] Step S44: Select another fault diagnosis path, calculate the fitness value, compare it with the previous calculation result, and sort it according to the comparison result to obtain the sorting result.

[0090] Then, the Grey Wolf optimization algorithm is used to calculate the fitness value of all fault diagnosis paths for the same possible fault. The fitness value is a key indicator to measure the quality of a fault diagnosis path. It comprehensively considers factors such as the effectiveness, accuracy, and efficiency of the fault diagnosis path in dealing with various possible faults. For example, for a fault diagnosis path, if it can quickly and accurately detect faults with high probability and serious impact, and is relatively economical in terms of resource consumption, then its fitness value will be relatively high. Conversely, if a fault diagnosis path has a high rate of missed detection or false detection and requires a lot of time and resources to detect faults, its fitness value will be low. After multiple comparisons, the paths are ranked according to the comparison results, and the fault diagnosis path with the highest fitness value has the highest priority.

[0091] Step S5: Perform fault diagnosis based on the sorting results. Select the fault diagnosis path from largest to smallest according to the sorting results to perform fault diagnosis.

[0092] Throughout the iterative calculation process, the Grey Wolf optimization algorithm fully leverages the advantages of swarm intelligence. Individuals within the Grey Wolf population collaborate and share information, working together towards the optimal solution. Fault diagnosis paths with higher fitness values ​​stand out and are given higher priority. In actual fault diagnosis work, once the system detects a potential fault, it will prioritize these fault diagnosis paths in descending order of fitness for investigation and diagnosis, quickly locating potential faults in the reaction flywheel. This greatly improves the efficiency and accuracy of fault diagnosis, effectively ensuring the stable operation of the entire system and reducing potential risks and losses caused by delays in fault diagnosis.

[0093] This invention discloses a fault diagnosis method for reaction flywheels. After collecting operational data from the reaction flywheel, a neural network model is used to process the pre-processed data, enabling rapid identification of potential faults and their probabilities. The method then identifies the components of the reaction flywheel used for fault diagnosis and maps them onto a virtual sphere. Several fault diagnosis paths are formed by connecting nodes on the virtual sphere's surface with their centers. A gray wolf optimization algorithm is then used to compare and rank these paths based on potential faults. Finally, a fault diagnosis path is selected based on the ranking results to diagnose the reaction flywheel fault, quickly locating the fault, improving fault diagnosis efficiency, and preventing disruption to the normal operation of the spacecraft.

[0094] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A fault diagnosis method for a reaction flywheel, characterized in that, Includes the following steps: Step S1: Collect the operating data of the reaction flywheel and preprocess the operating data; Step S2: Based on the preprocessed operating data, obtain the possible faults of the reaction flywheel and the probability of fault occurrence; Step S3: Construct the fault diagnosis path for the reaction flywheel; Step S4: The gray wolf optimization algorithm is used to process the fault diagnosis path based on the possible faults and the probability of fault occurrence, and the fault diagnosis path is sorted. Step S5: Perform fault diagnosis based on the sorting results; The specific steps of step S3 are as follows: Step S31: Obtain several diagnostic components for fault diagnosis of the reaction flywheel; Step S32: Construct a virtual sphere and uniformly map the diagnostic components onto the surface of the virtual sphere; Step S33: Starting from the center of the virtual sphere, randomly select at least one diagnostic component on the surface of the virtual sphere and connect them sequentially to form several fault diagnosis paths.

2. The fault diagnosis method for a reaction flywheel according to claim 1, characterized in that, The specific steps of step S1 include: Step S11: Collect real-time speed, motor current, bearing temperature and vibration data of the reaction flywheel; Step S12: Perform data cleaning, noise reduction, and feature extraction on the collected real-time speed, motor current, bearing temperature, and vibration data.

3. The fault diagnosis method for a reaction flywheel according to claim 1, characterized in that, The specific steps of step S2 include: Step S21: Construct a neural network model and train it to map the normal range of the reaction flywheel operation data into the neural network model; Step S22: Input the running data into the neural network model, and let the neural network model process the running data; Step S23: The neural network model outputs the possible faults of the reaction flywheel and the probability of the fault occurring.

4. The fault diagnosis method for a reaction flywheel according to claim 3, characterized in that, The specific steps for constructing and training the neural network model in step S21 are as follows: acquire historical data of the reaction flywheel, which includes historical operating data and fault records of the reaction flywheel. Take the historical operating data and fault records as a set of data and divide them into a training set and a test set in a 7:3 ratio. Train the neural network using the training set and test it using the test set after training. Once the test accuracy reaches the set threshold, the trained neural network model is obtained.

5. A fault diagnosis method for a reaction flywheel according to claim 1, characterized in that, Step S3 further includes: Step S34: Based on the possible faults of the reaction flywheel, remove irrelevant parts from the fault diagnosis path.

6. The fault diagnosis method for a reaction flywheel according to claim 1, characterized in that, The specific steps of step S4 include: Step S41: Select potential faults based on the probability of occurrence and the severity of the fault; Step S42: Initialize the gray wolf population and set the number of iterations; Step S43: Randomly select a fault diagnosis path and calculate the fitness value based on the possible faults; Step S44: Select another fault diagnosis path, calculate the fitness value, compare it with the previous calculation result, and sort it according to the comparison result to obtain the sorting result.

7. A fault diagnosis method for a reaction flywheel according to claim 6, characterized in that, The specific steps of step S41 include: Step S411: Assess and quantify the severity of the potential impact of the failure; Step S412: Assign values ​​to the probability of failure and the quantified degree of impact, and sum them up. Step S413: Select the possible faults from largest to smallest based on the summation result.

8. A fault diagnosis method for a reaction flywheel according to claim 6, characterized in that, The number of iterations in step S42 is the number of fault diagnosis paths.

9. A fault diagnosis method for a reaction flywheel according to claim 1 or 6, characterized in that, The specific steps of step S5 are as follows: select the fault diagnosis path from largest to smallest according to the sorting results and perform fault diagnosis.

Images