An aircraft mission sensor multi-process domain test data fusion information pickup method
By combining cross-entropy and linear and nonlinear correlation coefficients, information is extracted from multi-process domain test data of aircraft mission sensors, solving the problem of comprehensive accuracy evaluation in existing technologies and achieving efficient feature extraction with low complexity.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHENGDU AIRCRAFT INDUSTRY GROUP
- Filing Date
- 2023-06-20
- Publication Date
- 2026-07-14
AI Technical Summary
Existing information acquisition methods cannot comprehensively evaluate the accuracy of multi-process domain test data while ensuring low time complexity, cannot simultaneously reflect linear and nonlinear correlations, and are subject to noise.
The cross-entropy method is used to extract information from small sample process domain data, and linear and nonlinear correlation coefficients are combined to process large sample process domain data to construct a feature set of multiple process domains, ensuring the correlation between features and target parameters.
It achieves comprehensive information extraction from multi-process domain test data with low time complexity, improves the correlation between features and target parameters, and is suitable for efficient mining of heterogeneous data.
Smart Images
Figure CN116842356B_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of intelligent analysis of sensor test data, specifically involving a method for acquiring information fusion from multi-process domain test data of aircraft mission sensors. Background Technology
[0002] By analyzing the test data of mission sensors in various processes of aircraft production, it is possible to infer whether the accuracy of mission sensors during flight meets the factory specifications. This has significant practical value and meaning for improving aircraft flight test efficiency and ensuring a high first-pass yield.
[0003] However, since the original aircraft mission sensor test data involves multiple process domains, and the test data dimensions and sample sizes of different process domains vary greatly, the quality of the test data used for mission sensor accuracy assessment is reduced. The original test data cannot be directly used for accuracy assessment, and redundant test data will significantly increase the storage requirements and computational complexity of accuracy assessment.
[0004] Therefore, it is necessary to extract information from the test data of the task sensors across multiple process domains in order to extract effective test data information. Currently, commonly used information extraction methods often rely on some relevant metric variable to measure the correlation between various features of the original data and the target parameters, thereby selecting features with high correlation to the target parameters to form a new dataset to replace the original data for data mining. This type of method can reflect the correlation between the target and multidimensional data features and has advantages such as high efficiency and good generalization ability.
[0005] However, since current common information picking methods only select data based on a single relevant metric variable, they cannot adapt to the massive amount of computation of massive data while ensuring that the selected feature set has both good linear and nonlinear correlations, which has certain limitations.
[0006] Analysis revealed that the relationship between the parameters of the aircraft multi-process domain test data and flight accuracy is complex, exhibiting both linear and nonlinear correlation characteristics. Furthermore, it is susceptible to noise from the external environment, and the data dimensions and sample sizes vary across different process domains. To explore the correlation between each feature and the target parameter more comprehensively while maintaining low time complexity, it is necessary to use different relevant measurement variables for information extraction in each process domain based on the characteristics of the data in each process domain. Summary of the Invention
[0007] The purpose of this invention is to address the problems existing in the prior art, namely, that current information acquisition methods, when applied to multi-process domain test data, cannot comprehensively cover target parameter-related information while maintaining low time complexity. Therefore, this invention proposes a method for fusing information acquisition from multi-process domain test data of aircraft mission sensors.
[0008] To achieve the above-mentioned technical effects, the technical solution of this application is as follows:
[0009] A method for acquiring information from multi-process domain test data fusion of aircraft mission sensors includes the following steps:
[0010] First, the process domains are divided into small-sample process domain data and large-sample process domain data based on the amount of data. For small-sample process domain data, information is extracted based on cross-entropy. For large-sample process domain data, a combination of linear and nonlinear correlation coefficients is used to extract information from the large-sample data, resulting in more comprehensive information extraction results. Finally, based on this, a complete feature set with high correlation between multiple process domains and target parameters is obtained, realizing information extraction from multiple process domain data.
[0011] Furthermore, the specific steps of a method for acquiring multi-process domain test data fusion information from aircraft mission sensors are as follows:
[0012] Step 1: Obtain multi-domain data;
[0013] Step 2: Divide the multi-domain data into small-sample process domain data and large-sample process domain data;
[0014] Step 3: For small sample process domain data, the information picking method based on cross-entropy is used. Cross-entropy is used to characterize the similarity between two distributions. Cross-entropy is characterized by calculating the joint distribution density of features and target parameters. Based on the cross-entropy, the correlation between features and target parameters is judged, the correlation coefficient is calculated, and a threshold is set to filter out features that meet the correlation threshold requirements, thereby performing information picking.
[0015] Step four: For large sample process domain data, first calculate the covariance and standard deviation of each feature and the target parameter, and select the features that have a high linear correlation with the target variable; then analyze the independence between each feature and the target parameter, and select non-linearly correlated features based on this, so as to obtain more comprehensive information picking results.
[0016] Step 5: Summarize the feature sets obtained from small sample process domain data and the feature sets obtained from large sample process domain data to obtain a complete feature set with high correlation between multiple process domains and target parameters, thereby realizing information acquisition from multiple process domain data.
[0017] Furthermore, the standard for dividing multi-domain data into small-sample process domain data and large-sample process domain data in step two is a method already existing in this field.
[0018] Furthermore, the specific process for step three is as follows:
[0019] 1) Assume a task system obtains a dataset after multi-process domain testing. D represents the number of process domains. After preprocessing the dataset, the preprocessed dataset is obtained as follows: The dataset of the i-th process domain after preprocessing is denoted as:
[0020]
[0021] The target parameter is denoted as:
[0022]
[0023] Where n represents the number of samples in the task system, and m represents the number of features in this process domain;
[0024] 2) Perform joint distribution density calculation: Discretize the features and target variables in a two-dimensional space and represent them using a scatter plot. Divide the current two-dimensional space into x-squares along the X-axis and y-axis, respectively. Calculate the distribution density based on the distribution of scatter points within each square. and The joint distribution density is given by the following formula:
[0025]
[0026] Where, n ij Let n be the number of points that fall in the cell located in the i-th row and j-th column. iY Let n be the sum of the points that fall in all cells located in the i-th row. Xj Let m be the sum of the points that fall in all cells located in column j, and m be the total number of points.
[0027] The correlation between the characteristics and the target parameters is calculated from the joint distribution density. Recorded as:
[0028]
[0029] Adjust the position and number of lines so that most points are distributed in a few grid cells. This will yield the maximum cross-entropy, allowing you to calculate the maximum correlation coefficient. As shown in the following formula:
[0030]
[0031] Where B is the upper bound of the searchable grid, which controls the complexity of the correlations that can be detected.
[0032] By picking parameters related to flight accuracy in the ground test process domain using correlation coefficients, and assuming that the number of parameters picked in each process domain is uniformly represented by m', and n represents the number of samples in the mission system, the final data after picking information in the i-th process domain can be expressed as:
[0033]
[0034] Furthermore, the preprocessing of the dataset in step 1) includes: removing duplicate values and outliers and filling in missing values.
[0035] Furthermore, step four specifically involves:
[0036] For large sample process domain data with a large number of samples in some process domains, a method combining linear and nonlinear correlation coefficients is used to pick up information from large sample process domain data. First, the large sample process domain data is preprocessed. The preprocessed flight data is the same as the expression of formula (1).
[0037] Assume the k-th characteristic expression of the process domain data is The flight accuracy sequence is the same as the expression in formula (2);
[0038] First, calculate the linear correlation coefficient between process domain parameters and flight accuracy. The closer the correlation coefficient is to 1, the stronger the linear correlation between the two variables. Based on the correlation coefficient, select the appropriate parameter items according to the actual work needs.
[0039] Then, calculate the nonlinear correlation coefficient between the process domain parameters and flight accuracy. When the distance correlation coefficient is 0, it means that the two parameters are independent of each other. The larger the distance correlation coefficient, the stronger the distance correlation between the two parameters. Based on the correlation coefficient, select the parameter that meets the requirements according to the actual work needs.
[0040] Furthermore, preprocessing for large-sample process domain data datasets includes: removing duplicate and outlier values and filling in missing values.
[0041] Furthermore, the specific formula for calculating the linear correlation coefficient between process domain parameters and flight accuracy is as follows:
[0042]
[0043] Where E(·) is the expected value or mean, D(·) is the variance, the square root is the standard deviation, and Cov(x) is the mean. i ,y i Let x be a random variable. i y i The covariance.
[0044] Furthermore, the specific formula for calculating the nonlinear correlation coefficient between process domain parameters and flight accuracy is as follows:
[0045]
[0046] in, and They are respectively:
[0047]
[0048]
[0049]
[0050] Similarly, calculate and
[0051] Furthermore, step five specifically involves: combining the relevant process domain parameter terms selected from the linear and nonlinear coefficients to ultimately obtain the output of the information picking technique as follows:
[0052]
[0053] The advantages of this application are:
[0054] To achieve information extraction from heterogeneous test data, this patent proposes a method that first processes the data format to construct a sample sequence consistent with the characteristics of the target parameters, and then extracts information from the heterogeneous data based on the joint probability distribution density. The designed method can accurately extract features highly correlated with flight accuracy from heterogeneous data. Data format transformation makes more features consistent with the characteristics of the target parameters, meeting the input requirements of the information extraction model and facilitating a more thorough mining of the correlation information between heterogeneous data and target parameters. The information extraction method based on the joint probability distribution has high accuracy and can simultaneously extract linear and nonlinear correlation features related to the target parameters, making it suitable for small sample data. Mechanistic analysis shows that the parameters extracted have a high degree of correlation.
[0055] This invention proposes an information extraction method for small-sample-large-sample fusion, designing a low-complexity method with comprehensive feature information coverage for samples of different sizes. The research object of this invention is test information across multiple process domains, enabling information extraction from data of different data types. Attached Figure Description
[0056] Figure 1 Flowchart of information retrieval implementation method. Detailed Implementation
[0057] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0058] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
[0059] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
[0060] In the description of this application, it should be noted that the terms "upper," "vertical," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product is in use, or the orientation or positional relationship commonly understood by those skilled in the art. They are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on this application. In addition, the terms "first," "second," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0061] In the description of this application, it should also be noted that, unless otherwise expressly specified and limited, the terms "set," "install," and "connect" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.
[0062] Example 1
[0063] A method for acquiring information from multi-process domain test data fusion of aircraft mission sensors includes the following steps:
[0064] To ensure high parameter relevance and low computational complexity in information extraction for different process domains, a fusion method is adopted based on data characteristics to achieve information extraction from multiple process domains. First, the process domains are divided into small-sample and large-sample process domain data based on the amount of data. For small-sample process domain data, the overall computational cost is relatively low, and information extraction is achieved based on cross-entropy. For large-sample process domain data, which contains a large number of sample points in some process domains, this method results in high computational complexity, affecting information extraction efficiency. In comparison, linear and nonlinear correlation coefficients have significantly lower time complexity. However, due to the complex and diverse correlation between test data and target parameters, the feature sets extracted solely by linear or nonlinear correlation coefficients are insufficient. Therefore, a method combining linear and nonlinear correlation coefficients is considered for information extraction from large-sample data to obtain more comprehensive information extraction results. Finally, based on this, a complete feature set with high correlation between multiple process domains and target parameters is obtained, achieving information extraction from multiple process domains. To achieve information extraction from heterogeneous test data, this patent proposes a method that first processes the data format to construct a sample sequence consistent with the characteristics of the target parameters, and then extracts information from the heterogeneous data based on the joint probability distribution density. The designed method can accurately extract features highly correlated with flight accuracy from heterogeneous data. Data format transformation makes more features consistent with the characteristics of the target parameters, meeting the input requirements of the information extraction model and facilitating a more thorough mining of the correlation information between heterogeneous data and target parameters. The information extraction method based on the joint probability distribution has high accuracy and can simultaneously extract linear and nonlinear correlation features related to the target parameters, making it suitable for small sample data. Mechanistic analysis shows that the parameters extracted have a high degree of correlation.
[0065] Example 2
[0066] like Figure 1 As shown, the specific steps of a method for acquiring multi-process domain test data fusion information from aircraft mission sensors are as follows:
[0067] Step 1: Obtain multi-domain data;
[0068] Step 2: Divide the multi-domain data into small-sample process domain data and large-sample process domain data;
[0069] Step 3: For small sample process domain data, an information picking method based on cross-entropy is used. Cross-entropy is used to characterize the similarity between two distributions. It is characterized by calculating the joint distribution density of features and target parameters, and the correlation between features and target parameters is judged based on cross-entropy. The correlation coefficient is calculated, and a threshold is set to filter out features that meet the correlation threshold requirements, thereby picking information. The underlying idea is that if there is a correlation between two variables, then a grid should be drawn on the scatter plot of the variables so that most data points are concentrated in a few cells of the grid. By searching for the most suitable grid and combining it with the joint distribution density, a wide range of correlations between variables can be found.
[0070] Step 4: For large-sample process domain data, some process domains contain a large number of time series samples. First, calculate the covariance and standard deviation of each feature and the target parameter, and select the features that have a high linear correlation with the target variable. Then, analyze the independence between each feature and the target parameter, and select non-linearly correlated features based on this, so as to obtain more comprehensive information picking results.
[0071] Step 5: Summarize the feature sets obtained from small sample process domain data and the feature sets obtained from large sample process domain data to obtain a complete feature set with high correlation between multiple process domains and target parameters, thereby realizing information acquisition from multiple process domain data.
[0072] Example 3
[0073] like Figure 1 As shown, step one involves obtaining multi-domain data;
[0074] Step 2: Divide the multi-domain data into small-sample process domain data and large-sample process domain data;
[0075] Step 3: For small sample process domain data, an information picking method based on cross-entropy is used. Cross-entropy is used to characterize the similarity between two distributions. It is characterized by calculating the joint distribution density of features and target parameters, and the correlation between features and target parameters is judged based on cross-entropy. The correlation coefficient is calculated, and a threshold is set to filter out features that meet the correlation threshold requirements, thereby picking information. The underlying idea is that if there is a correlation between two variables, then a grid should be drawn on the scatter plot of the variables so that most data points are concentrated in a few cells of the grid. By searching for the most suitable grid and combining it with the joint distribution density, a wide range of correlations between variables can be found.
[0076] Step 4: For large-sample process domain data, some process domains contain a large number of time series samples. First, calculate the covariance and standard deviation of each feature and the target parameter, and select the features that have a high linear correlation with the target variable. Then, analyze the independence between each feature and the target parameter, and select non-linearly correlated features based on this, so as to obtain more comprehensive information picking results.
[0077] Step 5: Summarize the feature sets obtained from small sample process domain data and the feature sets obtained from large sample process domain data to obtain a complete feature set with high correlation between multiple process domains and target parameters, thereby realizing information acquisition from multiple process domain data.
[0078] The standard for dividing multi-domain data into small-sample process domain data and large-sample process domain data in step two is an existing method in this field.
[0079] The specific process for step three is as follows:
[0080] 1) Assume a task system obtains a dataset after multi-process domain testing. D represents the number of process domains. After preprocessing the dataset, the preprocessed dataset is obtained as follows: The dataset of the i-th process domain after preprocessing is denoted as:
[0081]
[0082] The target parameter is denoted as:
[0083]
[0084] Where n represents the number of samples in the task system, and m represents the number of features in this process domain;
[0085] 2) Perform joint distribution density calculation: Discretize the features and target variables in a two-dimensional space and represent them using a scatter plot. Divide the current two-dimensional space into x-squares along the X-axis and y-axis, respectively. Calculate the distribution density based on the distribution of scatter points within each square. and The joint distribution density is given by the following formula:
[0086]
[0087] Where, n ij Let n be the number of points that fall in the cell located in the i-th row and j-th column. iY Let n be the sum of the points that fall in all cells located in the i-th row. Xj Let m be the sum of the points that fall in all cells located in column j, and m be the total number of points.
[0088] The correlation between the characteristics and the target parameters is calculated from the joint distribution density. Recorded as:
[0089]
[0090] Adjust the position and number of lines so that most points are distributed in a few grid cells. This will yield the maximum cross-entropy, allowing you to calculate the maximum correlation coefficient. As shown in the following formula:
[0091]
[0092] Where B is the upper bound of the searchable grid, which controls the complexity of the correlations that can be detected.
[0093] By picking parameters related to flight accuracy in the ground test process domain using correlation coefficients, and assuming that the number of parameters picked in each process domain is uniformly represented by m', and n represents the number of samples in the mission system, the final data after picking information in the i-th process domain can be expressed as:
[0094]
[0095] The preprocessing of the dataset in step 1) includes: removing duplicate values and outliers and filling in missing values.
[0096] Step four is as follows:
[0097] For large sample process domain data with a large number of samples in some process domains, a method combining linear and nonlinear correlation coefficients is used to pick up information from large sample process domain data. First, the large sample process domain data is preprocessed. The preprocessed flight data is the same as the expression of formula (1).
[0098] Assume the k-th characteristic expression of the process domain data is The flight accuracy sequence is the same as the expression in formula (2);
[0099] First, calculate the linear correlation coefficient between process domain parameters and flight accuracy. The specific formula for calculating the linear correlation coefficient between process domain parameters and flight accuracy is as follows:
[0100]
[0101] Where E(i) is the expected value or mean, D(·) is the variance, the square root is the standard deviation, and Cov(x) is the mean. i ,y i Let x be a random variable. i y i The covariance.
[0102] The closer the correlation coefficient is to 1, the stronger the linear correlation between the two variables. Based on the correlation coefficient, select the appropriate parameter items according to the actual work needs.
[0103] Then, the nonlinear correlation coefficient between the process domain parameters and flight accuracy is calculated. The specific formula for calculating the nonlinear correlation coefficient between the process domain parameters and flight accuracy is as follows:
[0104]
[0105] in, and They are respectively:
[0106]
[0107]
[0108]
[0109] Similarly, calculate and
[0110] When the distance correlation coefficient is 0, it means that the two parameters are independent of each other. The larger the distance correlation coefficient, the stronger the distance correlation between the two parameters. Based on the correlation coefficient, select the parameter that meets the requirements according to the actual work needs.
[0111] Preprocessing for large-sample process domain data datasets includes: removing duplicate and outlier values and filling in missing values.
[0112] Step five specifically involves: combining the relevant process domain parameter terms selected for linear and nonlinear coefficients to ultimately obtain the output of the information picking technique.
[0113] Example 4
[0114] (1) In this case, flight altitude is used as the target parameter. Select test data from two process domains as input. in This is a small sample process domain containing 66 features; the data dimension is 1812*66. This is a small-sample process domain containing 88 features, with a data dimension of 7248*88. The dataset obtained by information gathering is used as the output.
[0115] (2) Test data of small sample process domains With target parameters Substituting into the "small sample data information picking" section of this patent, the correlation coefficients between each feature and the target parameter in the test data are obtained through formulas (3) to (6). Different thresholds are set according to the needs of subsequent mining tasks. Feature combinations with correlation coefficients greater than the thresholds are selected to obtain the final output dataset. The output results are shown in Table 7-1.
[0116] Table 7-1 Results of small sample process domain test data acquisition
[0117]
[0118]
[0119] (3) Large sample process domain test data With target parameters Substituting into the "large sample data information picking" section of this patent, the correlation coefficients between each feature and the target parameter in the test data are obtained through formula (7). Different thresholds are set according to the needs of subsequent mining tasks. Feature combinations with correlation coefficients greater than the thresholds are selected to obtain the final output dataset. The output results are shown in Table 7-2.
[0120] Table 7-2 Results of linear feature information extraction from large sample process domain test data
[0121]
[0122] Large sample process domain test data With target parameters Substituting into the "large sample data information picking" section of this patent, the correlation coefficients between each feature and the target parameter in the test data are obtained through formulas (8) to (12). Different thresholds are set according to the needs of subsequent mining tasks. Feature combinations with correlation coefficients greater than the thresholds are selected to obtain the final output dataset. The output results are shown in Table 7-3.
[0123] Table 7-3 Results of nonlinear feature information extraction from large sample process domain test data
[0124]
[0125] According to the set threshold, the linear feature information picking results and the nonlinear feature information picking results are merged to obtain the final information picking results of the large sample process domain test data.
[0126] (4) Based on the mechanism analysis, the information picking methods for both large sample process domain test data and small sample process domain test data can reduce the dimensionality of the original data while retaining the main relevant information of the target parameters, which proves the effectiveness of the method proposed in this invention.
Claims
1. A method for acquiring information from multi-process domain test data fusion of aircraft mission sensors, characterized in that: Includes the following steps: First, the process domains are divided into small-sample process domain data and large-sample process domain data based on the amount of data. For small-sample process domain data, information is extracted based on cross-entropy. For large-sample process domain data, a combination of linear and nonlinear correlation coefficients is used to extract information from the large-sample data, resulting in more comprehensive information extraction results. Finally, based on this, a complete feature set with high correlation between multiple process domains and target parameters is obtained, realizing information extraction from multiple process domain data. The specific steps are as follows: Step 1: Obtain multi-domain data; Step 2: Divide the multi-domain data into small-sample process domain data and large-sample process domain data; Step 3: For small sample process domain data, the information picking method based on cross-entropy is used. Cross-entropy is used to characterize the similarity between two distributions. Cross-entropy is characterized by calculating the joint distribution density of features and target parameters. Based on the cross-entropy, the correlation between features and target parameters is judged, the correlation coefficient is calculated, and a threshold is set to filter out features that meet the correlation threshold requirements, thereby performing information picking. Step four: For large sample process domain data, first calculate the covariance and standard deviation of each feature and the target parameter, and select the features that have a high linear correlation with the target variable; then analyze the independence between each feature and the target parameter, and select non-linearly correlated features based on this, so as to obtain more comprehensive information picking results. Step 5: Summarize the feature sets obtained from small sample process domain data and the feature sets obtained from large sample process domain data to obtain a complete feature set with high correlation between multiple process domains and target parameters, thereby realizing information acquisition from multiple process domain data.
2. The method for acquiring multi-process domain test data fusion information of aircraft mission sensors according to claim 1, characterized in that: The specific process for step three is as follows: 1) Assume a task system obtains a dataset after multi-process domain testing. , Given the number of process domains, after preprocessing the dataset, the preprocessed dataset is as follows: Then the first one after preprocessing i The dataset for each process domain is denoted as: （1） The target parameter is denoted as: （2） in, n This indicates the number of samples in the task system. m This indicates the number of features in this process domain; 2) Perform joint distribution density calculation: Discretize the features and target variables in a two-dimensional space and represent them using a scatter plot. Divide the current two-dimensional space into x-squares and y-squares along the X and Y axes, respectively. Calculate the distribution density based on the distribution of scatter points within each square. and The joint distribution density is given by the following formula: （3） in, Let be the number of points that fall in the cell located in row i and column j. The sum of the points that fall in all cells of the i-th row. Let m be the sum of the points that fall in all cells located in column j, and m be the total number of points. The correlation between the characteristics and the target parameters is calculated from the joint distribution density. , denoted as: （4） Adjust the position and number of lines so that most points are distributed in a few grid cells. This will yield the maximum cross-entropy, allowing you to calculate the maximum correlation coefficient. As shown in the following formula: （5） in The upper bound of the searchable grid controls the complexity of the detectable correlations; The parameters related to flight accuracy in the ground test process domain are picked up using correlation coefficients. It is assumed that the number of parameters picked up in each process domain is uniformly represented by [missing information]. express, n This represents the number of samples in the task system, ultimately yielding the [number of samples]. i The data obtained after process domain information is captured can be expressed as: （6）。 3. The method for acquiring multi-process domain test data fusion information of aircraft mission sensors according to claim 2, characterized in that: The preprocessing of the dataset in step 1) includes: removing duplicate values and outliers and filling in missing values.
4. The method for acquiring multi-process domain test data fusion information of aircraft mission sensors according to claim 2, characterized in that: Step four is as follows: For large sample process domain data with a large number of time series in some process domains, a method combining linear and nonlinear correlation coefficients is used to pick up information from large sample process domain data. First, the large sample process domain data is preprocessed. The preprocessed flight data is the same as the expression of formula (1). Assuming the process domain data is in the first... The characteristic expressions are as follows The flight accuracy sequence is the same as the expression in formula (2); First, calculate the linear correlation coefficient between process domain parameters and flight accuracy. The closer the correlation coefficient is to 1, the stronger the linear correlation between the two variables. Based on the correlation coefficient, select the appropriate parameter items according to the actual work needs. Then, calculate the nonlinear correlation coefficient between the process domain parameters and flight accuracy. When the distance correlation coefficient is 0, it means that the two parameters are independent of each other. The larger the distance correlation coefficient, the stronger the distance correlation between the two parameters. Based on the correlation coefficient, select the parameter that meets the requirements according to the actual work needs.
5. The method for acquiring multi-process domain test data fusion information of aircraft mission sensors according to claim 4, characterized in that: Preprocessing for large-sample process domain data datasets includes: removing duplicate and outlier values and filling in missing values.
6. A method for acquiring multi-process domain test data fusion information from aircraft mission sensors according to claim 4, characterized in that: The specific formula for calculating the linear correlation coefficient between process domain parameters and flight accuracy is as follows: （7） in, For mathematical expectation or mean, The variance is expressed as the square root of the standard deviation. For random variables , The covariance.
7. A method for acquiring multi-process domain test data fusion information from aircraft mission sensors according to claim 6, characterized in that: The specific formula for calculating the nonlinear correlation coefficient between process domain parameters and flight accuracy is as follows: （8） in, , , and They are respectively: （9） （10） （11） Similarly, calculate and .
8. The method for acquiring multi-process domain test data fusion information of aircraft mission sensors according to claim 7, characterized in that: Step five specifically involves: combining the relevant process domain parameter terms selected from the linear and nonlinear coefficients to obtain the final output of the information picking technique. （12）。