Method for determining source location of flight vibration data and method for determining vibration test conditions

Abstract

The application provides a flight vibration data source position judgment method and a vibration test condition determination method. The method comprises the following steps: obtaining historical flight vibration data, historical flight profile parameters and historical flight attitude parameters of a new aircraft or a similar aircraft; performing segmented analysis on the historical flight vibration data to obtain vibration acceleration root mean square of each time period; performing resampling on the historical flight profile parameters and the historical flight attitude parameters to obtain a feature data table; using the feature data table to establish a data source position judgment model through machine learning and performing screening and elimination, taking the model with the highest training precision as a final data source position judgment model; and using the final data source position judgment model to judge the source position of dirty data. The technical scheme of the application solves the technical problem that the vibration environment condition cannot be accurately determined due to missing or error of data source labels of dirty data in flight test of an aircraft in the prior art.

Description

Technical Field

[0001] This invention relates to the field of environmental adaptability technology, and in particular to a method for determining the source location of flight vibration data and a method for determining vibration test conditions. Background Technology

[0002] During aircraft flight, vibration environmental measurements are frequently taken. The data can be used to determine vibration environmental conditions or analyze problems encountered during flight tests. However, poor data retention is a common problem, primarily manifested in inaccurate vibration data labeling, such as missing or incorrect installation locations, or incorrect orientations. This resulting "dirty" data causes significant inconvenience in later use and can even affect the accuracy of subsequent work. Therefore, substantial manpower and time are invested in verification and inspection in practice, yet such problems cannot be completely avoided. Consequently, there is an urgent need to utilize computer-aided methods to determine the source location of vibration data and automatically complete data labeling; however, currently, no such methods exist. Summary of the Invention

[0003] This invention provides a method for determining the source location of flight vibration data and a method for determining vibration test conditions, which can solve the technical problem in the prior art where the data source label of dirty data in aircraft flight tests is missing or incorrect, resulting in the inability to accurately determine the vibration environment conditions.

[0004] According to one aspect of the present invention, a method for determining the location of flight vibration data source is provided, the method comprising:

[0005] S1, acquire historical flight vibration data, historical flight profile parameters and historical flight attitude parameters of the new or similar aircraft;

[0006] S2, according to the preset time step, the historical flight vibration data of each vibration measurement position under the same flight stage are segmented, and the root mean square of vibration acceleration for each time period is calculated;

[0007] S3, based on the root mean square of vibration acceleration corresponding to each vibration measurement position, the historical flight profile parameters and historical attitude parameters are resampled by linear interpolation. The root mean square of vibration acceleration corresponding to each vibration measurement position, the resampled historical flight profile parameters, and the resampled historical attitude parameters are integrated into a feature data table.

[0008] S4, randomly group the data in the feature data table to obtain training data and validation data;

[0009] S5 utilizes training data to establish multiple data source location judgment models through machine learning, with each data source location judgment model having a corresponding training accuracy.

[0010] S6. Use the verification data to verify the established multiple data source location judgment models to obtain the verification accuracy of each data source location judgment model. For each data source location judgment model, determine whether its training accuracy and verification accuracy meet the requirements at the same time. If not, go to S7; if yes, go to S8.

[0011] S7: Repeat S4 to S6 directly or regroup the data in the feature data table and then repeat S4 to S6. After each round of repeating S4 to S6, re-determine whether the training accuracy and verification accuracy of the data source location judgment model simultaneously meet the requirements. If yes, proceed to S8. If no, determine whether the number of rounds of repeating S4 to S6 is greater than or equal to the preset number of rounds. If no, proceed to the next round. If yes, remove the data source location judgment model.

[0012] S8, use the data source location determination model as the alternative data source location determination model;

[0013] S9 selects the candidate data source location judgment model with the highest training accuracy as the final data source location judgment model, and substitutes the new flight profile parameters and new attitude parameters of the new aircraft into the final data source location judgment model to obtain the data source location judgment result of the flight vibration data.

[0014] Furthermore, the criteria for identifying similar aircraft include similar aerodynamic shape, similar aircraft structure, or similar flight attitude angle, overload, and dynamic pressure variation range covering the new aircraft.

[0015] Furthermore, in S6, the requirement for training accuracy is that the training accuracy is greater than or equal to the expected value, and the requirement for verification accuracy is that the verification accuracy is not less than a preset percentage of the training accuracy.

[0016] Furthermore, regrouping the data in the feature data table includes:

[0017] S71, randomly select data that account for a preset percentage of all data in the feature data table from the previous training data as new verification data;

[0018] S72 combines the previous round of model verification data with the remaining training data from S71 to form new training data.

[0019] Furthermore, regrouping the data in the feature data table includes:

[0020] S71' Group the data in the feature data table according to the magnitude of each feature dataset. The data with the lowest magnitude a% is grouped into group A, the data with the highest magnitude b% is grouped into group B, and the remaining data from 100% to a% to b% is grouped into group C.

[0021] S72': After removing duplicates from all A-group data sets obtained in S71', A1-group data is formed; after removing duplicates from B-group data sets, B1-group data is formed; after removing duplicates from all C-group data sets, C1-group data is formed.

[0022] S73', randomly select c% from the data in group A1 and record it as group A2; after deleting the data in group B1 that are duplicates of group A2, randomly select d% from the remaining data in group B1 and record it as group B2; randomly select e% from the data in group C1 and record it as group C2.

[0023] S74' takes the union of the data from groups A2, B2, and C2 as the new validation data, and uses the data from the feature data table after removing the new validation data as the new training data.

[0024] Furthermore, after S7 and before S8, the method also includes reducing the feature parameters of the data source location judgment model when the training accuracy and verification accuracy of the data source location judgment model simultaneously meet the requirements.

[0025] Furthermore, the reduction of feature parameters for the data source location determination model includes:

[0026] Remove one or more feature datasets from the feature data table to obtain a reduced feature data table;

[0027] The data in the reduced feature data table are randomly grouped to obtain reduced training data and reduced validation data;

[0028] By using reduced training data, a data source location determination model is trained through machine learning to obtain a data source location determination model with reduced parameters and the corresponding training accuracy.

[0029] The reduced validation data is used to validate the data source location judgment model after parameter reduction to obtain the corresponding validation accuracy. It is determined whether the validation accuracy and training accuracy of the data source location judgment model after parameter reduction still meet the requirements. If yes, the feature parameters corresponding to the removed feature dataset are removed from the data source location judgment model. If no, the feature parameters corresponding to the removed feature dataset are retained in the data source location judgment model.

[0030] Furthermore, after S8 and before S9, the method also includes:

[0031] Repeat steps S2 to S8 using vibration data from one or more other flight phases to obtain a candidate data source location judgment model for vibration data in each flight phase.

[0032] The model with the highest accuracy among the candidate data source location determination models corresponding to vibration data from all flight phases is selected as the final data source location determination model.

[0033] Furthermore, both historical and new flight profile parameters include dynamic pressure and flight phases, and both historical and new attitude parameters include X-axis overload, Y-axis overload, Z-axis overload, roll angle, pitch angle, yaw angle, combustion chamber pressure, and fuel flow.

[0034] According to another aspect of the present invention, a method for determining vibration test conditions is provided, wherein the method utilizes the aforementioned method for determining the source location of aircraft flight vibration data to obtain the vibration test conditions for each vibration location during each flight phase.

[0035] This invention provides a method for determining the source location of flight vibration data and a method for determining vibration test conditions. This method, for the first time, proposes a specific approach to classify the vibration of new models or new flight profiles using flight profile parameters, attitude parameters, and measured vibration data from similar models. It can accurately determine the data source location of vibration data during supersonic vehicle flight, and further, can summarize the corresponding vibration test conditions based on vibration data from different data source locations at different stages. This fills the gap in data identification methods for determining the vibration environment of supersonic vehicles. This method can be extended to the cleaning of vibration data from various subsonic and supersonic vehicles. Compared with existing technologies, this invention solves the technical problem of inaccurate determination of vibration environment conditions due to missing or incorrect data source labels in dirty flight test data. Attached Figure Description

[0036] The accompanying drawings, which form part of this specification, are provided to further illustrate embodiments of the invention and, together with the textual description, explain the principles of the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0037] Figure 1 A flowchart illustrating a method for determining the source location of flight vibration data according to a specific embodiment of the present invention is shown. Detailed Implementation

[0038] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0040] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0041] According to a specific embodiment of the present invention, a method for determining the location of flight vibration data source is provided, the method comprising:

[0042] S1, acquire historical flight vibration data, historical flight profile parameters and historical flight attitude parameters of the new or similar aircraft;

[0043] S2, according to the preset time step, the historical flight vibration data of each vibration measurement position under the same flight stage are segmented, and the root mean square of vibration acceleration for each time period is calculated;

[0044] S3, based on the root mean square of vibration acceleration corresponding to each vibration measurement position, the historical flight profile parameters and historical attitude parameters are resampled by linear interpolation. The root mean square of vibration acceleration corresponding to each vibration measurement position, the resampled historical flight profile parameters, and the resampled historical attitude parameters are integrated into a feature data table.

[0045] S4, randomly group the data in the feature data table to obtain training data and validation data;

[0046] S5 utilizes training data to establish multiple data source location judgment models through machine learning, with each data source location judgment model having a corresponding training accuracy.

[0047] S6. Use the verification data to verify the established multiple data source location judgment models to obtain the verification accuracy of each data source location judgment model. For each data source location judgment model, determine whether its training accuracy and verification accuracy meet the requirements at the same time. If not, go to S7; if yes, go to S8.

[0048] S7: Repeat S4 to S6 directly or regroup the data in the feature data table and then repeat S4 to S6. After each round of repeating S4 to S6, re-determine whether the training accuracy and verification accuracy of the data source location judgment model simultaneously meet the requirements. If yes, proceed to S8. If no, determine whether the number of rounds of repeating S4 to S6 is greater than or equal to the preset number of rounds. If no, proceed to the next round. If yes, remove the data source location judgment model.

[0049] S8, use the data source location determination model as the alternative data source location determination model;

[0050] S9 selects the candidate data source location judgment model with the highest training accuracy as the final data source location judgment model, and substitutes the new flight profile parameters and new attitude parameters of the new aircraft into the final data source location judgment model to obtain the data source location judgment result of the flight vibration data.

[0051] In this embodiment of the invention, the criteria for determining similar aircraft in S1 include similar aerodynamic shape, similar aircraft structure, or flight attitude angles, overload, and dynamic pressure variation ranges covering the new aircraft. If no existing aircraft has flight attitude angles, overload, and dynamic pressure variation ranges that can completely cover the new aircraft, the one with the largest coverage range can be selected. For vibration prediction of the engine operating section, only measured vibration data (historical flight vibration data) of models with the same engine type can be used. When the historical data of this model or similar models has already been classified by source location, vibration data with missing data source location labels or incorrect labels (hereinafter referred to as dirty data) can be classified according to the differences in vibration data characteristics at different locations. In addition, in this invention, both historical flight profile parameters and new flight profile parameters include dynamic pressure and flight stage, and both historical flight attitude parameters and new attitude parameters include X-axis overload, Y-axis overload, Z-axis overload, roll angle, pitch angle, heading angle, combustion chamber pressure, and fuel flow. The preset time step for analyzing historical vibration data in S2 is determined according to the actual situation, for example, 2s to 5s as a time period. In S3, when resampling historical flight profile parameters and historical flight attitude parameters, the resampling time corresponds to the arithmetic mean of each vibration analysis time period. In S5, as a specific embodiment of the present invention, MATLAB is used for machine learning, with the data source location as the learning target. Various classification methods, such as decision trees, discriminant analysis, Naive Bayes classifiers, support vector machines, nearest neighbor classifiers, and ensemble classifiers, are used to learn and establish various data source location judgment models (all in functional form). Furthermore, the preset number of rounds is determined according to actual conditions, and will not be detailed in this invention.

[0052] This configuration provides a method for determining the source location of flight vibration data. This method, for the first time, proposes a specific approach to classify the vibration of new models or new flight profiles using flight profile parameters, attitude parameters, and measured vibration data from similar models. It can accurately determine the data source location of vibration data during supersonic vehicle flight, and further, summarize the corresponding vibration test conditions based on vibration data from different data source locations at different stages. This fills the gap in data identification methods for determining the vibration environment of supersonic vehicles. This method can be extended to the cleaning of vibration data from various subsonic and supersonic vehicles. Compared with existing technologies, the technical solution of this invention can solve the technical problem in existing technologies where missing or incorrect data source labels for dirty flight test data prevent accurate determination of vibration environment conditions.

[0053] Furthermore, in S6, the requirement for training accuracy is that the training accuracy is greater than or equal to the expected value, and the requirement for verification accuracy is that the verification accuracy is not less than a preset percentage of the training accuracy. That is, it simultaneously determines whether the training accuracy of the data source location determination model is greater than or equal to the expected value, and whether the verification accuracy is not less than a preset percentage of the training accuracy. Only when the training accuracy is greater than or equal to the expected value and the verification accuracy is not less than a preset percentage of the training accuracy can it be considered that the training accuracy and verification accuracy of the data source location determination model simultaneously meet the requirements. The expected value and preset percentage are determined based on the actual situation, and will not be detailed in detail in this invention.

[0054] Furthermore, in one embodiment of the present invention, regrouping the data in the feature data table includes:

[0055] S71, randomly select data that account for a preset percentage of all data in the feature data table from the previous training data as new verification data;

[0056] S72 combines the previous round of model verification data with the remaining training data from S71 to form new training data.

[0057] In another embodiment of the invention, regrouping the data in the feature data table includes:

[0058] S71' Group the data in the feature data table according to the magnitude of each feature dataset. The data with the lowest magnitude a% is grouped into group A, the data with the highest magnitude b% is grouped into group B, and the remaining data from 100% to a% to b% is grouped into group C.

[0059] S72': After removing duplicates from all A-group data sets obtained in S71', A1-group data is formed; after removing duplicates from B-group data sets, B1-group data is formed; after removing duplicates from all C-group data sets, C1-group data is formed.

[0060] S73', randomly select c% from the data in group A1 and record it as group A2; after deleting the data in group B1 that are duplicates of group A2, randomly select d% from the remaining data in group B1 and record it as group B2; randomly select e% from the data in group C1 and record it as group C2.

[0061] S74' takes the union of the data from groups A2, B2, and C2 as the new validation data, and uses the data from the feature data table after removing the new validation data as the new training data.

[0062] The values ​​of a%, b%, c%, d%, and e% are determined based on actual conditions and will not be detailed here. These two methods can obtain training data that is distributed as evenly as possible, improving the accuracy of the data source location determination model. Of course, those skilled in the art can also use other grouping methods for regrouping, and this invention does not impose further limitations.

[0063] Furthermore, in this embodiment of the invention, after S7 and before S8, the method further includes reducing the feature parameters of the data source location judgment model when the training accuracy and verification accuracy of the data source location judgment model simultaneously meet the requirements.

[0064] Specifically, reducing the feature parameters of the data source location determination model includes:

[0065] Remove one or more feature datasets from the feature data table to obtain a reduced feature data table;

[0066] The data in the reduced feature data table are randomly grouped to obtain reduced training data and reduced validation data;

[0067] By using reduced training data, a data source location determination model is trained through machine learning to obtain a data source location determination model with reduced parameters and the corresponding training accuracy.

[0068] The reduced validation data is used to validate the data source location judgment model after parameter reduction to obtain the corresponding validation accuracy. It is determined whether the validation accuracy and training accuracy of the data source location judgment model after parameter reduction still meet the requirements. If yes, the feature parameters corresponding to the removed feature dataset are removed from the data source location judgment model. If no, the feature parameters corresponding to the removed feature dataset are retained in the data source location judgment model.

[0069] By reducing the parameters of the data source location determination model in this way, a simplified model can be obtained, thereby improving the model's versatility.

[0070] To further improve the accuracy of the data source location determination model, in this embodiment of the invention, after S8 and before S9, the method further includes:

[0071] Repeat steps S2 to S8 using vibration data from one or more other flight phases to obtain a candidate data source location judgment model for vibration data in each flight phase.

[0072] The model with the highest accuracy among the candidate data source location determination models corresponding to vibration data from all flight phases is selected as the final data source location determination model.

[0073] To gain a further understanding of the present invention, the following description is provided in conjunction with... Figure 1 The method for determining the source location of flight vibration data according to the present invention will be described in detail.

[0074] Step 1: For a new type of supersonic aircraft, it is necessary to determine the flight vibration environment. If a similar model has been tested, or if this model has been tested and measured data is available, then the measured vibration data (historical flight vibration data), flight profile data (historical flight profile parameters), and attitude data (historical flight attitude parameters) of this model are used as the basic data for vibration prediction.

[0075] Step 2: Divide the measured vibration data at each measurement location into segments (each time segment can be 2s to 5s), and analyze the root mean square acceleration value for each time segment.

[0076] Step 3: Data Integration. Based on the root mean square of vibration acceleration corresponding to each vibration measurement location, the historical flight profile parameters and historical attitude parameters are resampled using linear interpolation. The sampling time corresponds to the arithmetic mean of each vibration analysis time period in Step 2.

[0077] Step 4: Machine Learning. 90% of the integrated data records are used as training data for machine learning, and 10% are used as validation data for model verification. The data records used for machine learning are imported into a machine learning tool (MATLAB in this example). Using the data source location as the learning objective, various classification methods are employed, including decision trees, discriminant analysis, Naive Bayes classifiers, support vector machines, nearest neighbor classifiers, and ensemble classifiers, to establish corresponding data source location judgment models (all in functional form).

[0078] Step 5: Accuracy Judgment of Model Verification. When using model verification data for location determination, if the accuracy of the determination (model verification accuracy) is lower than the accuracy of the determination (training accuracy) obtained by machine learning on the training data by a certain degree (this value is determined based on computing power, classification cycle, and the importance of subsequent analysis; for example, it can be determined to be more than 20%, that is, the model verification accuracy is more than 20% lower than the training accuracy), it indicates that the training data used for machine learning is not representative enough. In this case, adjust the data grouping according to Step 6; otherwise, proceed to Step 7.

[0079] Step 6: Repeat steps 4 and 5 directly, or repeat steps 4 and 5 after regrouping the data. Data regrouping can be done using the following methods:

[0080] Method a:

[0081] Step 1: Randomly select 10% of the data from the previous training data as the validation data.

[0082] Step 2: Combine the verification data from the previous round with the remaining training data from Step 1 to form new training data;

[0083] Method b:

[0084] Step 1: Grouping single-feature data. Divide the entire dataset into three groups according to the magnitude of each feature: the lowest 10% as group A, the highest 10% as group B, and the rest as group C;

[0085] Step 2: Constructing the full feature data grouped dataset. Take the A set of data from all feature datasets, keeping only one duplicate record, to form group A1. Take the B set of data from all feature datasets, keeping only one duplicate record, to form group B1. After removing data from groups A1 and B1 from all data, group C1 is formed (at this point, there may be duplicate data in A1 and B1).

[0086] Step 3: Randomly select 10% from group A1 and label it group A2. Delete the data that is duplicated from group B1 and label it group B2. Then randomly select 10% from group C1 and label it group C2.

[0087] Step 4: Determine the validation data and training data. Take the union of groups A2, B2, and C2 to form the validation data. Remove the validation data from all data and use the remaining data as the training data.

[0088] Step 7: Reduce training parameters. When the training accuracy of the data source location judgment model reaches over 90% and the verification accuracy is not less than 20% of the training accuracy, the training parameters are reduced to achieve the prediction accuracy with fewer training parameters. In this embodiment, X-axis overload is combined with angle parameters (C31, C32, and C33 combinations of roll angle, pitch angle, and yaw angle). Prediction models with training accuracy over 90% and verification accuracy not less than 20% of the training accuracy are sorted from high to low training accuracy. When the training accuracy is the same, the model with fewer training parameters is ranked higher.

[0089] Step 8: Determine the data source location judgment model. The prediction model with the highest training accuracy will be used as the final data source location judgment model.

[0090] Step 9: Determine the data source location. Substitute the parameters of the new flight profile and the new attitude parameters corresponding to the dirty data of the new model into the determined data source location judgment model to obtain the data source location judgment result of the dirty data.

[0091] According to another aspect of the present invention, a method for determining vibration test conditions is provided. This method utilizes the aforementioned method for determining the source location of aircraft flight vibration data to obtain the vibration test conditions for each vibration location during each flight phase. Specifically, based on the data source location determination result, the root mean square of vibration acceleration at the same location is statistically summarized for different phases, and the statistical summary result serves as the vibration test conditions for the corresponding region at that location during that phase. The statistical method can employ the maximum value envelope method or the normal one-sided tolerance upper limit method in QJ 20207—2012 "Requirements for Measurement and Processing of Environmental Data of Flight Missiles", or refer to GJB186. Furthermore, the data source location label of dirty data can be corrected to support further mining and use of the flight data record.

[0092] This approach provides a method for determining vibration test conditions. Since the aforementioned method for determining the source location of flight vibration data is the first to propose a specific method for classifying the vibration of a new model or flight profile using similar model flight profile parameters, attitude parameters, and measured vibration data, it can accurately determine the data source location of vibration data during supersonic vehicle flight. Furthermore, it can summarize the corresponding vibration test conditions based on vibration data from different data source locations at different stages, filling the gap in data identification methods for determining the vibration environment of supersonic vehicles. Therefore, applying this method to the determination of vibration test conditions can significantly improve the accuracy of supersonic vehicle vibration tests.

[0093] In summary, this invention provides a method for determining the source location of flight vibration data and a method for determining vibration test conditions. This method, for the first time, proposes a specific approach to classify the vibration of new models or new flight profiles using flight profile parameters, attitude parameters, and measured vibration data from similar models. It can accurately determine the data source location of vibration data during supersonic vehicle flight, and further, can summarize the corresponding vibration test conditions based on vibration data from different data source locations at different stages. This fills the gap in data identification methods for determining the vibration environment of supersonic vehicles. This method can be extended to the cleaning of vibration data from various subsonic and supersonic vehicles. Compared with existing technologies, the technical solution of this invention can solve the technical problem in existing technologies where missing or incorrect data source labels for dirty flight test data prevent accurate determination of vibration environment conditions.

[0094] For ease of description, spatial relative terms such as "above," "on top of," "on the upper surface of," "above," etc., are used herein to describe the spatial positional relationship of a device or feature as shown in the figures to other devices or features. It should be understood that spatial relative terms are intended to encompass different orientations in use or operation beyond the orientation of the device as described in the figures. For example, if the device in the figures were inverted, a device described as "above" or "on top of" other devices or structures would subsequently be positioned as "below" or "under" other devices or structures. Thus, the exemplary term "above" can include both "above" and "below." The device may also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatial relative descriptions used herein will be interpreted accordingly.

[0095] Furthermore, it should be noted that the use of terms such as "first" and "second" to define components is merely for the purpose of distinguishing the corresponding components. Unless otherwise stated, the above terms have no special meaning and therefore should not be construed as limiting the scope of protection of this invention.

[0096] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for determining the location of flight vibration data sources, characterized in that, The method includes: S1, acquire historical flight vibration data, historical flight profile parameters and historical flight attitude parameters of the new or similar aircraft; S2, according to the preset time step, the historical flight vibration data of each vibration measurement position under the same flight stage are segmented, and the root mean square of vibration acceleration for each time period is calculated; S3, based on the root mean square of vibration acceleration corresponding to each vibration measurement position, the historical flight profile parameters and the historical flight attitude parameters are resampled by linear interpolation. The root mean square of vibration acceleration corresponding to each vibration measurement position, the resampled historical flight profile parameters, and the resampled historical flight attitude parameters are integrated into a feature data table. S4, Randomly group the data in the feature data table to obtain training data and verification data; S5 utilizes training data to establish multiple data source location judgment models through machine learning, with each data source location judgment model having a corresponding training accuracy. S6. Use the verification data to verify the established multiple data source location judgment models to obtain the verification accuracy of each data source location judgment model. For each data source location judgment model, determine whether its training accuracy and verification accuracy meet the requirements at the same time. If not, go to S7; if yes, go to S8. S7. Repeat S4 to S6 directly or repeat S4 to S6 after regrouping the data in the feature data table. After each round of repeating S4 to S6, re-determine whether the training accuracy and verification accuracy of the data source location judgment model meet the requirements simultaneously. If yes, proceed to S8. If no, determine whether the number of rounds of repeating S4 to S6 is greater than or equal to the preset number of rounds. If no, proceed to the next round. If yes, remove the data source location judgment model. S8, use the data source location determination model as the alternative data source location determination model; S9 selects the candidate data source location judgment model with the highest training accuracy as the final data source location judgment model, and substitutes the new flight profile parameters and new attitude parameters of the new aircraft into the final data source location judgment model to obtain the data source location judgment result of the flight vibration data.

2. The method according to claim 1, characterized in that, The criteria for identifying similar aircraft include similar aerodynamic shape, similar aircraft structure, or similar flight attitude angle, overload and dynamic pressure variation range covering new aircraft.

3. The method according to claim 2, characterized in that, In S6, the requirement for training accuracy is that the training accuracy is greater than or equal to the expected value, and the requirement for verification accuracy is that the verification accuracy is not less than a preset percentage of the training accuracy.

4. The method according to any one of claims 1 to 3, characterized in that, Regrouping the data in the feature data table includes: S71, randomly select data that account for a preset percentage of all data in the feature data table from the previous training data as new verification data; S72 combines the previous round of model verification data with the remaining training data from S71 to form new training data.

5. The method according to any one of claims 1 to 3, characterized in that, Regrouping the data in the feature data table includes: S71' Group the data in the feature data table according to the magnitude of each feature dataset. The data with the lowest magnitude a% is grouped into group A, the data with the highest magnitude b% is grouped into group B, and the remaining data 100%-a%-b% is grouped into group C. S72', after removing duplicates from all A-group data sets obtained in S71', form A1-group data; after removing duplicates from the B-group data sets, form B1-group data; after removing duplicates from all C-group data sets, form C1-group data. S73', randomly select c% from the data in group A1 and record it as group A2; after deleting the data in group B1 that are duplicates of group A2, randomly select d% from the remaining data in group B1 and record it as group B2; randomly select e% from the data in group C1 and record it as group C2. S74' takes the union of the data from groups A2, B2, and C2 as the new validation data, and uses the data from the feature data table after removing the new validation data as the new training data.

6. The method according to claim 1, characterized in that, After S7 and before S8, the method further includes reducing the feature parameters of the data source location judgment model when the training accuracy and verification accuracy of the data source location judgment model simultaneously meet the requirements.

7. The method according to claim 6, characterized in that, Reducing the feature parameters of the data source location determination model includes: Remove one or more feature datasets from the feature data table to obtain a reduced feature data table; The data in the reduced feature data table are randomly grouped to obtain reduced training data and reduced validation data; By using reduced training data, a data source location determination model is trained through machine learning to obtain a data source location determination model with reduced parameters and the corresponding training accuracy. The reduced validation data is used to validate the data source location judgment model after parameter reduction to obtain the corresponding validation accuracy. It is determined whether the validation accuracy and training accuracy of the data source location judgment model after parameter reduction still meet the requirements. If yes, the feature parameters corresponding to the removed feature dataset are removed from the data source location judgment model. If no, the feature parameters corresponding to the removed feature dataset are retained in the data source location judgment model.

8. The method according to claim 1, characterized in that, The method further includes, after S8 and before S9: Repeat steps S2 to S8 using vibration data from one or more other flight phases to obtain a candidate data source location judgment model for vibration data in each flight phase. The model with the highest accuracy among the candidate data source location determination models corresponding to vibration data from all flight phases is selected as the final data source location determination model.

9. The method according to claim 8, characterized in that, Both historical and new flight profile parameters include dynamic pressure and flight phases. Both historical and new attitude parameters include X-axis overload, Y-axis overload, Z-axis overload, roll angle, pitch angle, heading angle, combustion chamber pressure, and fuel flow.

10. A method for determining vibration test conditions, characterized in that, The method uses the aircraft flight vibration data source location determination method according to any one of claims 1 to 9 to obtain the vibration test conditions for each vibration location in each flight phase.

Images