Intelligent and reliable spacecraft flywheel system safety assessment method

Abstract

The invention discloses an intelligent and reliable spacecraft flywheel system safety assessment method, and belongs to the technical field of spacecraft flywheel system safety assessment, and the method comprises the following steps: 1, screening input data; 2, mining a reference value; 3, setting a matching degree and an attribute weight threshold value; step 4, rule reduction; 5, model reasoning; step 6, optimizing the model; according to the method, intelligent and reliable safety assessment can be carried out on the spacecraft flywheel system in a complex interference environment, reasonable data are screened from the collected observation data, the reasonability and effectiveness of input data are ensured, reference values are mined by adopting a clustering algorithm, a matching degree threshold value is set, the generation of an over-low matching degree is avoided, and the safety of the spacecraft flywheel system is improved. According to the input information conversion method, the robustness of the input information conversion process is enhanced, meanwhile, the use of invalid rules is reduced by setting rule weight threshold values and reducing redundant rules with low weights, the complexity of the model is reduced, and the robustness and reliability of the model are further enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of safety assessment of a spacecraft flywheel system, and in particular relates to an intelligent and reliable safety assessment method for a spacecraft flywheel system. Background Art

[0002] With the continuous development of aerospace technology, the performance and reliability requirements of spacecraft are increasing day by day. As a key component of spacecraft attitude control, the safety assessment of spacecraft flywheel system is of vital importance. At present, the traditional safety assessment method of spacecraft flywheel system has certain limitations. On the one hand, the assessment means are relatively single, often relying only on limited sensor data for analysis, which is difficult to fully and accurately reflect the actual operating status of the flywheel system. On the other hand, the lack of intelligent assessment means makes it impossible to monitor and predict potential failure risks in real time. At the same time, due to the complex and changeable operating environment of spacecraft, the existing assessment methods are not robust and reliable enough when facing extreme working conditions and emergencies, and may not be able to issue warnings in a timely and effective manner. In order to meet the high reliability and long life operation requirements of spacecraft, there is an urgent need for an intelligent and reliable safety assessment method for spacecraft flywheel system, which can comprehensively utilize multi-source data to realize real-time monitoring, fault diagnosis and risk prediction of the flywheel system, and provide strong guarantee for the safe operation of spacecraft.

[0003] Therefore, establishing an intelligent and reliable safety assessment method for spacecraft flywheel systems has important practical significance. Summary of the invention

[0004] The purpose of the present invention is to propose an intelligent and reliable safety assessment method for a spacecraft flywheel system in order to solve the problems raised in the above background technology.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] An intelligent and reliable safety assessment method for a spacecraft flywheel system comprises the following steps:

[0007] Step 1: Screen input data. First, screen reliable input data.

[0008] Step 2: Reference value mining. After the input data screening in step 1 is completed, a clustering algorithm is used to mine reasonable reference values ​​in the safety assessment process.

[0009] Step 3: setting the matching degree and attribute weight thresholds, based on the filtered input data obtained in step 1 and the reasonable reference values ​​obtained in step 2;

[0010] Step 4: rule simplification: based on the matching degree and attribute weight thresholds set in step 3, simplifies the redundant rules with lower weights;

[0011] Step 5: Model reasoning: perform model reasoning based on the above steps;

[0012] Step six, model optimization, evaluates the model performance and optimizes it based on the model inference results.

[0013] As a further description of the above technical solution: In the step 1, x i,j represents the i-th data of the j-th input set, i=1,2,...,M,j=1,2,...,T, where x low and x top Represents x i,j The maximum and minimum values ​​that can be taken are the range of a reasonable interval. Data outside the interval are error data and are not selected into the experimental data set.

[0014] As a further description of the above technical solution: in step 2, the clustering algorithm process includes:

[0015] First, use the reference value provided by the expert as the initial cluster center and initialize the reference value:

[0016]

[0017] Among them, ci represents the i-th cluster, μ i represents the i-th cluster center, T represents the number of cluster centers;

[0018] Second, calculate the shortest distance from each data point to the current set of cluster centers as follows:

[0019]

[0020] Among them, d j and x j They represent the jth data point of the security assessment index data, M represents the total number of data points, and dist(x j ,u i ) represents the data point x j To cluster center u i distance;

[0021] Third, update the cluster assigned to each data point, calculated as follows:

[0022] c i =arg min dist(x j ,u i ) (3)

[0023] Among them, argmin represents the index of the minimum value;

[0024] Fourth, update the cluster center. The calculation formula is as follows:

[0025]

[0026] Fifth, calculate the minimum sum of squared errors in clustering:

[0027]

[0028] Where J represents the minimum sum of squared errors in clustering;

[0029] Repeat steps 2 to 5 until a certain criterion is met or the maximum number of iterations is reached. At this point, the obtained cluster center represents the optimized reference value:

[0030]

[0031] As a further description of the above technical solution: In step 3, the matching threshold is set:

[0032]

[0033] Wherein, a represents the matching degree threshold. When the matching degree exceeds a, the original matching degree is retained; otherwise, the matching degree is set to 0.

[0034] As a further description of the above technical solution: In the step 3, the simplification of redundant rules with smaller weights is achieved by setting a rule weight threshold:

[0035]

[0036] Among them, θ represents the tolerance threshold of rule weight.

[0037] As a further description of the above technical solution: In step 4, the reasonable range of attribute weights can be described as follows:

[0038] υ≤δ m ≤1,(m=1,...,M) (9)

[0039] Among them, υ is defined by the expert system as a reasonable threshold of attribute weight.

[0040] As a further description of the above technical solution: in step 5, the model reasoning process includes:

[0041] First, calculate the matching degree between the input information and the confidence rule, the matching degree between the i-th input and the k-th rule Obtained by the following formula:

[0042]

[0043] Second, calculate the activation weight of the kth rule:

[0044]

[0045] in, represents the normalized weight of the i-th attribute;

[0046] Third, the evidence reasoning parsing algorithm is used to generate the confidence of different evaluation results. The confidence and utility value of the nth evaluation result are calculated as follows:

[0047]

[0048] Fourth, calculate the expected utility value and get the evaluation result:

[0049]

[0050] Where μ(D n ) represents D n The utility value of A' represents the output vector in the actual system, S(·) represents a set of confidence distributions, and μ(S(A')) represents the final expected utility value. The final confidence distribution can be expressed as:

[0051] y={(D n ,β n ),n=1,...,N} (16).

[0052] As a further description of the above technical solution: in step 6, an improved projection covariance matrix adaptive evolution strategy optimization algorithm with robustness constraints is used for optimization, and the calculation formula of the optimization error is as follows:

[0053]

[0054] in, and They represent the actual output and expected output of the tth group of data respectively. The calculation method of the optimization objective function is as follows:

[0055]

[0056] The specific optimization process is as follows:

[0057] First, parameter optimization, the initial parameters are set to ω 0 =ψ 0 , the number of update iterations is set to G, and the initial search step size is set to ε 0 , the covariance matrix is ​​labeled C 0 , and the population size is ξ;

[0058] Second, the sampling operation, the sampling operation is described as follows:

[0059]

[0060] in, is the qth solution of the (t+1)th generation, ω is the overall average, ε represents the step constant, represents the normal distribution, C t Represents the covariance matrix of the t-th generation population;

[0061] Third, the projection operation is described as follows:

[0062]

[0063] Among them, v e Indicated in The number of variables in the solution, p = 1, 2, ..., N + 1, is The number of constants in A e =[1...1] 1×N A parameter vector representing the sampling operation;

[0064] Fourth, the selection operation is performed to update the average value, which is described as follows:

[0065]

[0066] in, is the qth solution output in the ξth solution of the (t+1)th generation;

[0067] Fifth, the adjustment operation, the covariance matrix is ​​updated, which is described as follows:

[0068]

[0069] Among them, c 1 and c 2 is the learning rate, p c is the evolution path of the covariance, and the evolution process is as follows:

[0070]

[0071] Sixth, the step size is updated as follows:

[0072]

[0073] Among them, c c is the backward timeline of the evolutionary path, c σ represents the enumerated evolution path vector, d σ represents the attenuation coefficient, Indicates p σ The expected length of represents the mathematical expectation, p σ Indicates enumerating evolution paths;

[0074] Seventh, termination criterion, after G rounds of optimization, the optimization is terminated and the optimization vector of the entire process is returned.

[0075] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0076] In the present invention, the present invention can perform intelligent and reliable safety assessment on the flywheel system of a spacecraft in a complex interference environment. By screening the input data, obviously abnormal or inaccurate data can be removed to ensure that subsequent assessments are based on high-quality data, avoid deviations in assessment results due to erroneous data, and lay the foundation for accurate assessments. By using a clustering algorithm to mine reference values ​​for historical data, past experience and data can be fully utilized to provide reasonable standards and references for current safety assessments, making the assessments more objective and accurate. By setting a matching threshold, an excessively low matching degree can be avoided, and the robustness of the input information conversion process is enhanced, making the assessment results more in line with actual conditions and needs. By removing redundant and unnecessary rules, the assessment process is simplified, and the assessment efficiency is improved. At the same time, the confusion and uncertainty that may be caused by complex rules are avoided. The model is optimized by an improved optimization algorithm with robustness constraints, so that it can better adapt to different situations and interferences, improve the anti-interference ability and robustness of the model, and ultimately obtain reliable safety assessment results to provide a basis for final decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 A flow chart of an intelligent and reliable safety assessment method for a spacecraft flywheel system proposed by the present invention;

[0078] Figure 2 A clustering algorithm flow chart of an intelligent and reliable spacecraft flywheel system safety assessment method proposed by the present invention;

[0079] Figure 3 The overall diagram of an intelligent and reliable spacecraft flywheel system safety assessment method proposed by the present invention. DETAILED DESCRIPTION

[0080] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0081] Please refer to the attached Figure 1 -Attached Figure 3The present invention provides a technical solution: an intelligent and reliable safety assessment method for a spacecraft flywheel system, comprising the following steps:

[0082] Step 1: Screen input data. First, screen reliable input data.

[0083] Step 2: Reference value mining. After the input data screening in step 1 is completed, a clustering algorithm is used to mine reasonable reference values ​​in the safety assessment process.

[0084] Step 3: setting the matching degree and attribute weight thresholds, based on the filtered input data obtained in step 1 and the reasonable reference values ​​obtained in step 2;

[0085] Step 4: rule simplification: based on the matching degree and attribute weight thresholds set in step 3, simplifies the redundant rules with lower weights;

[0086] Step 5: Model reasoning: perform model reasoning based on the above steps;

[0087] Step six, model optimization, evaluates the model performance and optimizes it based on the model inference results.

[0088] In step one, in step one, x i,j represents the i-th data of the j-th input set, i=1,2,...,M,j=1,2,...,T, where x low and x top Represents x i,j The maximum and minimum values ​​that can be taken are the range of a reasonable interval. Data outside the interval are error data and are not selected into the experimental data set.

[0089] In step 2, the clustering algorithm process includes:

[0090] First, use the reference value provided by the expert as the initial cluster center and initialize the reference value:

[0091]

[0092] Among them, c i represents the i-th cluster, μ i represents the i-th cluster center, T represents the number of cluster centers;

[0093] Second, calculate the shortest distance from each data point to the current set of cluster centers as follows:

[0094]

[0095] Among them, d j and x jThey represent the jth data point of the security assessment index data, M represents the total number of data points, and dist(x j ,u i ) represents the data point x j To cluster center u i distance;

[0096] Third, update the cluster assigned to each data point, calculated as follows:

[0097] c i = arg min dist(x j ,u i ) (3)

[0098] Among them, argmin represents the index of the minimum value;

[0099] Fourth, update the cluster center. The calculation formula is as follows:

[0100]

[0101] Fifth, calculate the minimum sum of squared errors in clustering:

[0102]

[0103] Where J represents the minimum sum of squared errors in clustering;

[0104] Repeat steps 2 to 5 until a certain criterion is met or the maximum number of iterations is reached. At this point, the obtained cluster center represents the optimized reference value:

[0105]

[0106] In step 3, set the matching threshold:

[0107]

[0108] Wherein, a represents the matching degree threshold. When the matching degree exceeds a, the original matching degree is retained; otherwise, the matching degree is set to 0.

[0109] In step 3, the redundant rules with smaller weights are simplified by setting the rule weight threshold:

[0110]

[0111] Among them, θ represents the tolerance threshold of rule weight.

[0112] In step 4, the reasonable range of attribute weights can be described as follows:

[0113] υ≤δ m ≤1,(m=1,...,M) (9)

[0114] Among them, υ is defined by the expert system as a reasonable threshold of attribute weight.

[0115] In step 5, the model inference process includes:

[0116] First, calculate the matching degree between the input information and the confidence rule, the matching degree between the i-th input and the k-th rule Obtained by the following formula:

[0117]

[0118] Second, calculate the activation weight of the kth rule:

[0119]

[0120] in, represents the normalized weight of the i-th attribute;

[0121] Third, the evidence reasoning parsing algorithm is used to generate the confidence of different evaluation results. The confidence and utility value of the nth evaluation result are calculated as follows:

[0122]

[0123] Fourth, calculate the expected utility value and get the evaluation result:

[0124]

[0125] Where μ(D n ) represents D n The utility value of A' represents the output vector in the actual system, S(·) represents a set of confidence distributions, and μ(S(A')) represents the final expected utility value. The final confidence distribution can be expressed as:

[0126] y={(D n ,β n ),n=1,...,N} (16).

[0127] In step six, an improved projection covariance matrix adaptive evolutionary strategy optimization algorithm with robust constraints is used for optimization. The calculation formula of the optimization error is as follows:

[0128]

[0129] in, and They represent the actual output and expected output of the tth group of data respectively. The calculation method of the optimization objective function is as follows:

[0130]

[0131] The specific optimization process is as follows:

[0132] First, parameter optimization, the initial parameters are set to ω 0 =ψ 0 , the number of update iterations is set to G, and the initial search step size is set to ε 0 , the covariance matrix is ​​labeled C 0 , and the population size is ξ;

[0133] Second, the sampling operation, the sampling operation is described as follows:

[0134]

[0135] in, is the qth solution of the (t+1)th generation, ω is the overall average, ε represents the step constant, represents the normal distribution, C t Represents the covariance matrix of the t-th generation population;

[0136] Third, the projection operation is described as follows:

[0137]

[0138] Among them, v e Indicated in The number of variables in the solution, p = 1, 2, ..., N + 1, is The number of constants in A e =[1...1] 1×N A parameter vector representing the sampling operation;

[0139] Fourth, the selection operation is performed to update the average value, which is described as follows:

[0140]

[0141] in, is the qth solution output in the ξth solution of the (t+1)th generation;

[0142] Fifth, the adjustment operation, the covariance matrix is ​​updated, which is described as follows:

[0143]

[0144] Among them, c 1 and c 2 is the learning rate, p c is the evolution path of the covariance, and the evolution process is as follows:

[0145]

[0146] Sixth, the step size is updated as follows:

[0147]

[0148] Among them, c c is the backward timeline of the evolutionary path, c σ represents the enumerated evolution path vector, d σ represents the attenuation coefficient, Indicates p σ The expected length of represents the mathematical expectation, p σ Indicates enumerating evolution paths;

[0149] Seventh, termination criterion, after G rounds of optimization, the optimization is terminated and the optimization vector of the entire process is returned.

[0150] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An intelligent and reliable safety assessment method for a spacecraft flywheel system, characterized in that: The following steps are involved: Step 1: Screen input data. First, screen reliable input data. Step 2: Reference value mining. After the input data screening in step 1 is completed, a clustering algorithm is used to mine reasonable reference values ​​in the safety assessment process. Step 3: setting the matching degree and attribute weight thresholds, based on the filtered input data obtained in step 1 and the reasonable reference values ​​obtained in step 2; Step 4: rule simplification: based on the matching degree and attribute weight thresholds set in step 3, simplifies the redundant rules with lower weights; Step 5: Model reasoning: perform model reasoning based on the above steps; Step six, model optimization, evaluates the model performance and optimizes it based on the model inference results.

2. An intelligent and reliable spacecraft flywheel system safety assessment method according to claim 1, characterized in that: In step 1, the selected reliable observation data can be expressed as x low ≤x i,j ≤x top .

3. The intelligent and reliable safety assessment method for a spacecraft flywheel system according to claim 1, characterized in that: In the step 1, x i,j represents the i-th data of the j-th input set, i=1,2,...,M,j=1,2,...,T, where x low and x top Represents x i,j The maximum and minimum values ​​that can be taken are the range of a reasonable interval. Data outside the interval are error data and are not selected into the experimental data set.

4. The intelligent and reliable spacecraft flywheel system safety assessment method according to claim 1, characterized in that: In the step 2, the clustering algorithm process includes: First, use the reference value provided by the expert as the initial cluster center and initialize the reference value: Among them, ci represents the i-th cluster, μ i represents the i-th cluster center, T represents the number of cluster centers; Second, calculate the shortest distance from each data point to the current set of cluster centers as follows: Among them, d j and x j They represent the jth data point of the security assessment index data, M represents the total number of data points, and dist(x j ,u i ) represents the data point x j To cluster center u i distance; Third, update the cluster assigned to each data point, calculated as follows: c i =arg min dist(x j ,u i ) (3) Among them, argmin represents the index of the minimum value; Fourth, update the cluster center. The calculation formula is as follows: Fifth, calculate the minimum sum of squared errors in clustering: Where J represents the minimum sum of squared errors in clustering; Repeat steps 2 to 5 until a certain criterion is met or the maximum number of iterations is reached. At this point, the obtained cluster center represents the optimized reference value:

5. The intelligent and reliable spacecraft flywheel system safety assessment method according to claim 1, characterized in that: In step 3, the matching threshold is set: Wherein, a represents the matching degree threshold. When the matching degree exceeds a, the original matching degree is retained; otherwise, the matching degree is set to 0.

6. The intelligent and reliable spacecraft flywheel system safety assessment method according to claim 1, characterized in that: In step 3, the redundant rules with smaller weights are simplified by setting the rule weight threshold: Among them, θ represents the tolerance threshold of rule weight.

7. The intelligent and reliable safety assessment method for a spacecraft flywheel system according to claim 1, characterized in that: In step 4, the reasonable range of attribute weights can be described as follows: υ≤δ m ≤1,(m=1,...,M) (9) Among them, υ is defined by the expert system as a reasonable threshold of attribute weight.

8. The intelligent and reliable spacecraft flywheel system safety assessment method according to claim 1, characterized in that: In step 5, the model reasoning process includes: First, calculate the matching degree between the input information and the confidence rule, the matching degree between the i-th input and the k-th rule It is obtained by the following formula: Second, calculate the activation weight of the kth rule: in, represents the normalized weight of the i-th attribute; Third, the evidence reasoning parsing algorithm is used to generate the confidence of different evaluation results. The confidence and utility value of the nth evaluation result are calculated as follows: Fourth, calculate the expected utility value and get the evaluation result: Where μ(D n ) represents D n The utility value of A' represents the output vector in the actual system, S(·) represents a set of confidence distributions, and μ(S(A')) represents the final expected utility value. The final confidence distribution can be expressed as: y={(D n ,β n ),n=1,…,N} (16)。 9. The intelligent and reliable spacecraft flywheel system safety assessment method according to claim 1, characterized in that: In step 6, an improved projection covariance matrix adaptive evolution strategy optimization algorithm with robustness constraints is used for optimization, and the calculation formula of the optimization error is as follows: in, and They represent the actual output and expected output of the tth group of data respectively. The calculation method of the optimization objective function is as follows: The specific optimization process is as follows: First, parameter optimization, the initial parameters are set to ω 0 =ψ 0 , the number of update iterations is set to G, and the initial search step size is set to ε 0 , the covariance matrix is ​​labeled C 0 , and the population size is ξ; Second, the sampling operation, the sampling operation is described as follows: in, is the qth solution of the (t+1)th generation, ω is the overall average, ε represents the step constant, represents the normal distribution, C t Represents the covariance matrix of the t-th generation population; Third, the projection operation is described as follows: Among them, v e Indicated in The number of variables in the solution, p = 1, 2, ..., N + 1, is The number of constants in A e =[1...1] 1×N A parameter vector representing the sampling operation; Fourth, the selection operation is performed to update the average value, which is described as follows: in, is the qth solution output in the ξth solution of the (t+1)th generation; Fifth, the adjustment operation, the covariance matrix is ​​updated, which is described as follows: Among them, c1 and c2 are learning rates, p c is the evolution path of the covariance, and the evolution process is as follows: Sixth, the step size is updated as follows: Among them, c c is the backward timeline of the evolutionary path, c σ represents the enumerated evolution path vector, d σ represents the attenuation coefficient, Indicates p σ The expected length of represents the mathematical expectation, p σ Indicates enumerating evolution paths; Seventh, termination criterion, after G rounds of optimization, the optimization is terminated and the optimization vector of the entire process is returned.

Images