Aircraft model parameter correction method and device based on real data

Abstract

The invention provides an aircraft model parameter correction method and device based on real data. The method comprises the following steps: acquiring a to-be-corrected flight path of a target aircraft model; inputting flight parameters extracted from the to-be-corrected flight path into a pre-constructed parameter correction model to obtain correction parameters output by the parameter correction model; performing reverse interpolation on the to-be-corrected flight path based on the correction parameter to obtain a parameter correction result; wherein the parameter correction model is obtained by training a flight segment with a label and a pre-stored initial parameter interpolation table based on a convolutional neural network; the flight segment with the label is obtained by performing data processing on a real flight data sample. And the accuracy of aircraft model parameter correction is improved.

Description

Technical Field

[0001] The present invention relates to the field of aviation technology, and in particular to a method and device for correcting aircraft model parameters based on real data. Background Art

[0002] In the intelligent aerospace field, the simulation model of aircraft plays a vital role. A complete simulation model can not only simulate the flight of aircraft to obtain relevant flight parameters, especially the aerodynamic parameters of some more complex actions during air combat, but also play a role in training real pilots and simulated pilots to conduct digital air combat. This not only saves training costs, but also obtains air combat game strategies combined with expert experience.

[0003] However, if the dynamic parameters of the aircraft model are not accurately estimated, the referenceability of the calculated results will be greatly affected. Therefore, the parameter correction of the aviation equipment simulation model is extremely important. Only by correcting the parameters of the simulation model can the simulated flight curve be as similar as possible to the real flight curve under the same operating conditions, and the simulation model can be truly meaningful and instructive. In addition, if the correction process of the model parameters can have a certain degree of universality, it can realize the transfer learning of the simulation model for the type B aircraft under the condition that the aircraft model of type A and the real flight data of type B aircraft are known.

[0004] In the existing technology, the parameter correction method for the aircraft model is mainly to solve the model parameter estimation problem through the least squares method. The disadvantage is that it lacks robustness, that is, it is extremely sensitive to the errors of the observation data, which affects the accuracy of the simulated flight equipment model. Summary of the invention

[0005] In order to solve the problems existing in the existing technology, the present invention provides a method and device for correcting aircraft model parameters based on real data, so as to improve the accuracy of aircraft model parameter correction.

[0006] The present invention provides a method for correcting aircraft model parameters based on real data, the method comprising:

[0007] Obtaining the flight trajectory to be corrected of the target aircraft model;

[0008] Inputting the flight parameters extracted from the flight trajectory to be corrected into a pre-built parameter correction model to obtain correction parameters output by the parameter correction model;

[0009] Performing reverse interpolation on the flight trajectory to be corrected based on the correction parameters to obtain a parameter correction result;

[0010] The parameter correction model is obtained by training based on a convolutional neural network using labeled flight segments and a pre-stored initial parameter interpolation table; the labeled flight segments are obtained after data processing of real flight data samples.

[0011] In some embodiments, the parameter correction result includes a corrected lift coefficient, drag coefficient, and thrust interpolation table.

[0012] In some embodiments, based on a convolutional neural network, training is performed using labeled flight segments and a pre-stored initial parameter interpolation table to obtain the parameter correction model, specifically including:

[0013] Acquire real flight data samples, and preprocess the real flight data samples to obtain a training data set; the training data set includes flight parameters of the real flight data samples and labels corresponding to the flight parameters;

[0014] The training data set is input into a pre-built convolutional neural network for training to obtain the parameter correction model.

[0015] In some embodiments, the real flight data samples are preprocessed to obtain a training data set, specifically including:

[0016] Obtaining flight parameters of real flight data samples and labels corresponding to the flight parameters;

[0017] Initializing the flight parameters according to a predetermined number of initialization times to obtain initialization parameters;

[0018] Performing noise processing on the initialization parameters according to a predetermined number of noise additions;

[0019] Perform length segmentation on the parameters obtained after noise processing;

[0020] The training data set is constructed based on the sliced ​​flight trajectory data segments and corresponding labels obtained after segmentation.

[0021] In some embodiments, constructing the training data set based on the segmented flight parameters and corresponding labels specifically includes:

[0022] Generate a fitting interpolation table based on the initial interpolation table through noise addition processing;

[0023] A large number of initial parameters are combined with the fitting interpolation table generated by multiple noise additions to obtain a large number of sliced ​​flight trajectory data segments with flight coefficient labels;

[0024] The training data set is constructed using the sliced ​​flight trajectory data segments and the corresponding flight coefficient labels.

[0025] In some embodiments, the flight parameters include a Mach number-lift coefficient interpolation table, a lift coefficient-drag coefficient interpolation table, an angle of attack-thrust interpolation table, flight position parameters, flight angle parameters and flight speed parameters, etc., a missile flight parameter interpolation table and a missile flight trajectory parameters.

[0026] The present invention also provides an aircraft model parameter correction device based on real data, the device comprising:

[0027] A data acquisition unit, used for acquiring the flight trajectory to be corrected of the target aircraft model;

[0028] A data processing unit, used for inputting the flight parameters extracted from the flight trajectory to be corrected into a pre-built parameter correction model to obtain correction parameters output by the parameter correction model;

[0029] A result generating unit, used for performing reverse interpolation on the flight trajectory to be corrected based on the correction parameters to obtain a parameter correction result;

[0030] The parameter correction model is obtained by training based on a convolutional neural network using labeled flight segments and a pre-stored initial parameter interpolation table; the labeled flight segments are obtained after data processing of real flight data samples.

[0031] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described above when executing the program.

[0032] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the method described above when executed by a processor.

[0033] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method described above is implemented.

[0034] The aircraft model parameter correction method and device based on real data provided by the present invention provide a flight trajectory to be corrected for a target aircraft model; the flight parameters extracted from the flight trajectory to be corrected are input into a pre-built parameter correction model to obtain correction parameters output by the parameter correction model; the flight trajectory to be corrected is reversely interpolated based on the correction parameters to obtain parameter correction results. The parameter correction method and device provided by the present invention respectively correct key parameters of the aircraft three-degree-of-freedom model, the missile three-degree-of-freedom model, and the missile guidance model for actual air combat data in an offline state, and adopt a convolutional neural network parameter correction model to improve the speed and accuracy of model parameter correction, and improve the accuracy of aircraft model parameter correction; in actual application, the model parameter correction method provided by the present invention can complete the parameter correction of a typical equipment model in a relatively short time, which helps to improve the accuracy of aircraft simulation and has engineering practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 One of the flow charts of the method for correcting aircraft model parameters based on real data provided by the present invention;

[0037] Figure 2 The second flowchart of the method for correcting aircraft model parameters based on real data provided by the present invention;

[0038] Figure 3 A schematic diagram of initializing slice flight parameters in a specific usage scenario;

[0039] Figure 4 Flow chart 3 of the method for correcting aircraft model parameters based on real data provided by the present invention;

[0040] Figure 5 for Figure 3 The flight data labeling result obtained using the nearest neighbor interpolation method in the specific usage scenario shown;

[0041] Figure 6 Flow chart 4 of the method for correcting aircraft model parameters based on real data provided by the present invention;

[0042] Figure 7 is a comparison diagram between the flight trajectory obtained according to the original interpolation table and the flight trajectory obtained according to the revised interpolation table;

[0043] Figure 8 Flow chart 5 of the method for correcting aircraft model parameters based on real data provided by the present invention;

[0044] Fig. 9 This is the root mean square error diagram of the convolutional neural network on the validation set after training;

[0045] Fig.10 Flow chart six of the method for correcting aircraft model parameters based on real data provided by the present invention;

[0046] Fig.11 A schematic diagram of the structure of the aircraft model parameter correction device based on real data provided by the present invention;

[0047] Fig.12 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0049] At present, with the popularization of artificial intelligence in data processing, it is increasingly integrated with engineering projects such as aerospace, electrical machinery and equipment. Based on the above-mentioned shortcomings of the prior art, in order to achieve rapid correction of typical flight equipment model parameters and effectively improve the accuracy of model parameter correction, the present invention proposes a method and device for correcting aircraft model parameters based on real data for massive flight data.

[0050] Please refer to Figure 1 , Figure 1 One of the flow charts of the method for correcting aircraft model parameters based on real data provided by the present invention;

[0051] In a specific embodiment, the method for correcting aircraft model parameters based on real data provided by the present invention comprises the following steps:

[0052] S110: Acquire the flight trajectory to be corrected of the target aircraft model; that is, for the target aircraft model, the original flight trajectory formed within a certain period of time is the flight trajectory to be corrected;

[0053] S120: Inputting the flight parameters extracted from the flight trajectory to be corrected into a pre-constructed parameter correction model to obtain correction parameters output by the parameter correction model; specifically, the flight parameters include a typical flight parameter interpolation table of an aircraft four-degree-of-freedom model (including a Mach number-lift coefficient interpolation table, a lift coefficient-drag coefficient interpolation table, and an angle of attack-thrust interpolation table), flight trajectory parameters (including flight position parameters, flight angle parameters, and flight speed parameters, etc.), a missile flight parameter interpolation table, and missile flight trajectory parameters.

[0054] S130: performing reverse interpolation on the flight trajectory to be corrected based on the correction parameters to obtain a parameter correction result; specifically, the parameter correction result may include a corrected lift coefficient, drag coefficient and thrust interpolation table;

[0055] The parameter correction model is obtained by training based on a convolutional neural network using labeled flight segments and a pre-stored initial parameter interpolation table; the labeled flight segments are obtained after data processing of real flight data samples.

[0056] When using the parameter correction model for detection and identification, first read in the real flight trajectory to be corrected stored in the correction folder. The real flight trajectory corresponds to the flight coefficient that is different from the initial interpolation table, that is, the lift coefficient, drag coefficient and thrust size that need to be corrected; call the stored best model parameters to test the model; divide the original trajectory data and store it in the corresponding folder; find the corresponding relationship between the real flight and the label, that is, the lift coefficient, drag coefficient and thrust, through the trained convolutional neural network, and obtain the corresponding parameters; use the parameters obtained according to the original flight trajectory and the convolution parameter correction neural network to perform reverse interpolation, and finally correct the original parameter interpolation table.

[0057] In this way, compared with the traditional aircraft model parameter correction method, the parameter correction method provided by the present invention can, on the one hand, improve the robustness of equipment model parameter correction, and on the other hand, greatly improve the efficiency of typical equipment model parameter correction; ultimately, a high-precision interpolation table of key parameters of the aircraft model can be obtained, providing a more accurate simulated flight trajectory, and providing effective technical support for the construction of a high-security vacuum combat confrontation simulation test system.

[0058] In some embodiments, based on a convolutional neural network, training is performed using labeled flight segments and a pre-stored initial parameter interpolation table to obtain the parameter correction model, specifically including:

[0059] Acquire real flight data samples, and preprocess the real flight data samples to obtain a training data set; the training data set includes flight parameters of the real flight data samples and labels corresponding to the flight parameters;

[0060] The training data set is input into a pre-constructed convolutional neural network for training to obtain the parameter correction model; that is, the convolutional neural network is trained using the training data set, and the parameter correction model after the training reaches the standard is stored.

[0061] Specifically, the convolutional neural network is trained based on the labeled flight segments obtained from data processing and the parameter interpolation table after initialization and noise addition, and the relationship between the lift coefficient, drag coefficient, thrust and flight trajectory is found through the convolutional neural network. After the training is completed, the real data is input into the trained model to obtain the corresponding most suitable lift coefficient, drag coefficient and thrust interpolation situation; on this basis, the corrected lift coefficient, drag coefficient and thrust interpolation table can be further inferred.

[0062] In some embodiments, Figure 2 As shown, the real flight data samples are preprocessed to obtain a training data set, specifically including the following steps:

[0063] S210: Acquire flight parameters of real flight data samples and labels corresponding to the flight parameters;

[0064] S220: Initializing the flight parameters according to a predetermined number of initialization times to obtain initialization parameters;

[0065] S230: performing noise processing on the initialization parameters according to a predetermined number of noise additions;

[0066] S240: performing length segmentation on the parameters obtained after the noise addition process;

[0067] S250: Constructing the training data set based on the sliced ​​flight trajectory data segments and corresponding labels obtained after segmentation.

[0068] Specifically, in the data preprocessing stage, it is necessary to first determine the number of initializations, that is, the number of flight initializations; determine the number of noise additions, that is, the number of noise additions for the initial parameters; determine the division length, that is, the length of the division of the long flight data; determine the number of training rounds, that is, the appropriate number of epochs for model training. Then, a new interpolation table is constructed based on the input initialization parameters and the number of noise additions to enrich the input model data; the flight data output by the software each time is divided according to the division length; at the same time, the flight parameters in this state, that is, the lift coefficient, drag coefficient and thrust size are labeled, and then the dimension is modified to form image format data, and finally encapsulated as neural network model trainable data. Name and store the data according to the current number of initializations and the current number of segmented segments.

[0069] The data set obtained through the above preprocessing process is used to train the convolutional neural network, so that the parameter correction model obtained through training can expand the model correction flight state range by initializing the model parameters multiple times during use, making the algorithm more versatile; at the same time, multiple parameter interpolation tables are used to add noise to construct new interpolation tables and enrich the input model data, thereby solving the problems of data collection difficulties and the difficulty of simulated flight trajectories to approach real trajectories, and improving the accuracy of parameter correction.

[0070] In step S250, the training data set is constructed based on the segmented flight parameters and corresponding labels, which specifically includes:

[0071] Generate a fitting interpolation table based on the initial interpolation table through noise addition processing;

[0072] A large number of initial parameters are combined with the fitting interpolation table generated by multiple noise additions to obtain a large number of sliced ​​flight trajectory data segments with flight coefficient labels;

[0073] The training data set is constructed using the sliced ​​flight trajectory data segments and the corresponding flight coefficient labels.

[0074] During the data preprocessing process, in order to ensure that the flight data contains valid data mining fragments, the file size of each piece of flight data is first screened; for example, all .txt files storing flight data are traversed, and files with a size less than 2500kb are removed, and only files larger than 2500kb are retained for subsequent equipment model parameter correction.

[0075] Secondly, since the operating environment of aircraft in the air is complex and changeable, there are often poor signals or signal disappearances. For the collected flight data, the missing values ​​must first be filled; the data is screened according to the time number, and if there are missing values, the linear interpolation method is used to fill the data. The specific operation is to select the flight data information of one time node before and after the data with time missing positions, take each feature as the dependent variable, take time as the independent variable, construct a linear interpolation function, substitute the missing time node into the interpolation function, and obtain the predicted value of the corresponding feature.

[0076] Since real flight data contains a lot of noise, denoising and smoothing operations are performed after missing value filling. Wavelet threshold denoising and hierarchical clustering smoothing methods are mainly used for data denoising and smoothing operations. When performing wavelet threshold denoising, the selection of wavelet basis should comprehensively consider the shape and smoothness of flight data, and select wavelet basis functions with similar shapes and smoothness. The selection of decomposition layers should ensure that noise is eliminated to the maximum extent while maintaining the image shape and trend.

[0077] At the same time, in order to avoid the imbalance of training weights of key parameters such as lift coefficient, drag coefficient and thrust due to differences in magnitude during model training, the method of adding scaling coefficients was adopted to unify multiple parameters to a certain order of magnitude and optimize the subsequent model training process.

[0078] In this way, in order to better approximate the actual interpolation of lift coefficient, drag coefficient and thrust, a sufficient number of fitting interpolation tables based on the initial interpolation table are generated by adding noise, and then a large number of differently initialized flight trajectory parameters are combined with the fitting interpolation tables generated by multiple noise additions, thereby obtaining a large number of sliced ​​flight trajectory data segments with typical flight coefficient labels, completing the data preparation stage for the subsequent convolutional neural network correction model.

[0079] For ease of understanding, the implementation process and technical effects of the method provided by the present invention are briefly described below using a specific usage scenario as an example.

[0080] like Figure 3 and Figure 4 As shown, Figure 3 A partial screenshot showing the label data list after slicing and storage. Figure 4 Demonstrates the use of the data processing method provided by the present invention to obtain Figure 3 The specific operation process of the data shown; according to Figure 4 As shown in the figure, firstly, the initial parameters of lift coefficient Cy, drag coefficient Cx, and thrust PF are taken as input, and random disturbances are added to them to achieve the purpose of increasing data samples. At the same time, model training requires data of different flight states (altitude, speed, and elevation angle). By adjusting the initial flight value and calling EXE multiple times, rich data is obtained to prepare for subsequent training. Since there are many calls and the initial value needs to be modified for each call, it will bring great time and personnel costs if this function cannot be automatically implemented through code. The automatic call of EXE is realized through the cmd command, and the dos window generated after the operation is closed. After adjusting the initial flight value and calling EXE multiple times, the network training data corresponding to different initial flight states are obtained, and the storage format is: xyz.npy. Among them, x represents the number of noise additions, y represents the number of initializations of the initial flight value, and z represents the number of segments of the flight process.

[0081] Figure 5 The data labeling results of the segmentation in the present invention are shown. The first part of each row is the storage location of the sliced ​​flight trajectory, and the second part is the corresponding key parameters, namely the lift coefficient, drag coefficient and thrust.

[0082] Figure 6 Demonstrates the use of the data processing method provided by the present invention to generate Figure 5The data shown in the figure is further added to the specific flow chart for model training and correction; First, in the process of labeling the segmented data, two-dimensional linear interpolation is required to obtain the lift coefficient, drag coefficient and thrust under the determined altitude, angle of attack and Mach number. In order to solve the problem of exceeding the limit, consider the data of the exceeding part, use the nearest neighbor interpolation method in two-dimensional linear interpolation, and replace the exceeding part with the nearest neighbor interpolation point, which is used as the label of the data segment.

[0083] Figure 7 It shows the effect comparison between the corrected flight trajectory obtained according to the convolutional neural network correction model in the present invention and the actual flight trajectory; wherein the R-Square between the two trajectories is 0.9997.

[0084] Figure 8 According to the present invention, Figure 7 Correction flow chart of the correction result; first, the initial value is modified multiple times and random disturbances are added to obtain the initialized flight parameters lift coefficient Cy, drag coefficient Cx, and thrust PF; then the corresponding flight trajectory data is obtained by cyclically calling the application program, and the data is segmented according to the set segment length, and the corresponding lift coefficient Cy, drag coefficient Cx, and thrust PF are obtained by two-dimensional linear interpolation calculation as labels to form network training data. Then, the lift coefficient Cy, drag coefficient Cx, and thrust PF parameters of the training data set are scaled to improve the data processing effect of the model.

[0085] The relationship between the lift coefficient, drag coefficient, thrust and model output is found through the convolutional neural network. The real data is input into the trained model to obtain the corresponding most suitable lift coefficient, drag coefficient and thrust interpolation. On this basis, the corrected lift coefficient, drag coefficient and thrust interpolation table can be further inferred.

[0086] According to a known flight trajectory data, it is first segmented, and then the lift coefficient, drag coefficient and thrust parameter of each segment under the corresponding state are obtained through the trained convolutional neural network. The altitude HH, Mach number Ma, power angle AA and the lift coefficient Cy, drag coefficient Cx and thrust parameter PF of the corresponding flight segment in the flight trajectory are taken as known, and the interpolation points under typical flight conditions in the interpolation table of the original lift coefficient Cy, drag coefficient Cx and thrust parameter PF are interpolated and corrected, and the interpolation table is modified; finally, the corrected interpolation table is brought into the application to re-describe the flight trajectory.

[0087] Fig. 9 The figure shows a root mean square error curve graph on the flight trajectory verification set obtained by the trained convolutional neural network in the present invention, and the root mean square error is controlled at about 0.39.

[0088] Fig.10 This is the overall flow chart of the convolutional neural network correction model involved in the typical equipment model parameter correction. After training according to the overall process, the correction model is obtained Fig.10 The error curve results.

[0089] Finally, the data preprocessing link and the overall model were encapsulated into interactive software to form a typical equipment model parameter correction platform, providing effective technical support for the construction of a high-security vacuum warfare confrontation simulation test system.

[0090] In the above specific implementation, the aircraft model parameter correction method based on real data provided by the present invention provides a flight trajectory to be corrected for the target aircraft model; the flight parameters extracted from the flight trajectory to be corrected are input into a pre-built parameter correction model to obtain the correction parameters output by the parameter correction model; the flight trajectory to be corrected is reversely interpolated based on the correction parameters to obtain the parameter correction result. The parameter correction method provided by the present invention corrects key parameters of the aircraft three-degree-of-freedom model, the missile three-degree-of-freedom model, and the missile guidance model respectively for actual air combat data in an offline state, and adopts a convolutional neural network parameter correction model to improve the speed and accuracy of model parameter correction, and improve the accuracy of aircraft model parameter correction; in actual application, the model parameter correction method provided by the present invention can complete the parameter correction of a typical equipment model in a relatively short time, which helps to improve the accuracy of aircraft simulation and has engineering practical value.

[0091] In addition to the above method, the present invention also provides an aircraft model parameter correction device based on real data, such as Fig.11 As shown, the device comprises:

[0092] The data acquisition unit 1110 is used to acquire the flight trajectory to be corrected of the target aircraft model;

[0093] The data processing unit 1120 is used to input the flight parameters extracted from the flight trajectory to be corrected into a pre-built parameter correction model to obtain correction parameters output by the parameter correction model;

[0094] A result generating unit 1130 is used to perform reverse interpolation on the flight trajectory to be corrected based on the correction parameters to obtain a parameter correction result;

[0095] The parameter correction model is obtained by training based on a convolutional neural network using labeled flight segments and a pre-stored initial parameter interpolation table; the labeled flight segments are obtained after data processing of real flight data samples.

[0096] In some embodiments, the parameter correction result includes a corrected lift coefficient, drag coefficient, and thrust interpolation table.

[0097] In some embodiments, based on a convolutional neural network, training is performed using labeled flight segments and a pre-stored initial parameter interpolation table to obtain the parameter correction model, specifically including:

[0098] Acquire real flight data samples, and preprocess the real flight data samples to obtain a training data set; the training data set includes flight parameters of the real flight data samples and labels corresponding to the flight parameters;

[0099] The training data set is input into a pre-built convolutional neural network for training to obtain the parameter correction model.

[0100] In some embodiments, the real flight data samples are preprocessed to obtain a training data set, specifically including:

[0101] Obtaining flight parameters of real flight data samples and labels corresponding to the flight parameters;

[0102] Initializing the flight parameters according to a predetermined number of initialization times to obtain initialization parameters;

[0103] Performing noise processing on the initialization parameters according to a predetermined number of noise additions;

[0104] Perform length segmentation on the parameters obtained after noise processing;

[0105] The training data set is constructed based on the sliced ​​flight trajectory data segments and corresponding labels obtained after segmentation.

[0106] In some embodiments, constructing the training data set based on the segmented flight parameters and corresponding labels specifically includes:

[0107] Generate a fitting interpolation table based on the initial interpolation table through noise addition processing;

[0108] A large number of initial parameters are combined with the fitting interpolation table generated by multiple noise additions to obtain a large number of sliced ​​flight trajectory data segments with flight coefficient labels;

[0109] The training data set is constructed using the sliced ​​flight trajectory data segments and the corresponding flight coefficient labels.

[0110] In some embodiments, the flight parameters include a Mach number-lift coefficient interpolation table, a lift coefficient-drag coefficient interpolation table, an angle of attack-thrust interpolation table, flight position parameters, flight angle parameters and flight speed parameters, etc., a missile flight parameter interpolation table and a missile flight trajectory parameters.

[0111] In the above specific implementation, the aircraft model parameter correction device based on real data provided by the present invention provides a method for obtaining the flight trajectory to be corrected of the target aircraft model; inputting the flight parameters extracted from the flight trajectory to be corrected into a pre-built parameter correction model to obtain the correction parameters output by the parameter correction model; and reversely interpolating the flight trajectory to be corrected based on the correction parameters to obtain the parameter correction result. The parameter correction device provided by the present invention corrects key parameters of the aircraft three-degree-of-freedom model, the missile three-degree-of-freedom model, and the missile guidance model respectively for actual air combat data in an offline state, and adopts a convolutional neural network parameter correction model to improve the speed and accuracy of model parameter correction, and improve the accuracy of aircraft model parameter correction; in actual application, the model parameter correction device provided by the present invention can complete the parameter correction of a typical equipment model in a relatively short time, which helps to improve the accuracy of aircraft simulation and has engineering practical value.

[0112] Fig.12 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.12 As shown, the electronic device may include: a processor 1210, a communication interface 1220, a memory 1230 and a communication bus 1240, wherein the processor 1210, the communication interface 1220, and the memory 1230 communicate with each other through the communication bus 1240. The processor 1210 may call the logic instructions in the memory 1230 to execute the above method.

[0113] In addition, the logic instructions in the above-mentioned memory 1230 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the part of the technical solution of the present invention that essentially contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instruction sets that enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0114] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can perform the above method.

[0115] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0116] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for correcting aircraft model parameters based on real data, characterized in that: The method comprises: Obtaining the flight trajectory to be corrected of the target aircraft model; Inputting the flight parameters extracted from the flight trajectory to be corrected into a pre-built parameter correction model to obtain correction parameters output by the parameter correction model; Performing reverse interpolation on the flight trajectory to be corrected based on the correction parameters to obtain a parameter correction result; The parameter correction model is obtained by training based on a convolutional neural network using labeled flight segments and a pre-stored initial parameter interpolation table; the labeled flight segments are obtained after data processing of real flight data samples.

2. The method for correcting aircraft model parameters based on real data according to claim 1, characterized in that: The parameter correction results include a corrected lift coefficient, drag coefficient and thrust interpolation table.

3. The method for correcting aircraft model parameters based on real data according to claim 1, characterized in that: Based on the convolutional neural network, the parameter correction model is obtained by training using labeled flight segments and a pre-stored initial parameter interpolation table, which specifically includes: Acquire real flight data samples, and preprocess the real flight data samples to obtain a training data set; the training data set includes flight parameters of the real flight data samples and labels corresponding to the flight parameters; The training data set is input into a pre-built convolutional neural network for training to obtain the parameter correction model.

4. The method for correcting aircraft model parameters based on real data according to claim 3, characterized in that: The real flight data samples are preprocessed to obtain a training data set, specifically including: Obtaining flight parameters of real flight data samples and labels corresponding to the flight parameters; Initializing the flight parameters according to a predetermined number of initialization times to obtain initialization parameters; Performing noise processing on the initialization parameters according to a predetermined number of noise additions; Perform length segmentation on the parameters obtained after noise processing; The training data set is constructed based on the sliced ​​flight trajectory data segments and corresponding labels obtained after segmentation.

5. The method for correcting aircraft model parameters based on real data according to claim 4, characterized in that: Based on the segmented flight parameters and corresponding labels, the training data set is constructed, specifically including: Generate a fitting interpolation table based on the initial interpolation table through noise addition processing; A large number of initial parameters are combined with the fitting interpolation table generated by multiple noise additions to obtain a large number of sliced ​​flight trajectory data segments with flight coefficient labels; The training data set is constructed using the sliced ​​flight trajectory data segments and the corresponding flight coefficient labels.

6. The method for correcting aircraft model parameters based on real data according to any one of claims 1 to 5, characterized in that: The flight parameters include a Mach number-lift coefficient interpolation table, a lift coefficient-drag coefficient interpolation table, an angle of attack-thrust interpolation table, flight position parameters, flight angle parameters, flight speed parameters, etc., a missile flight parameter interpolation table and missile flight trajectory parameters.

7. An aircraft model parameter correction device based on real data, characterized in that: The device comprises: A data acquisition unit, used for acquiring the flight trajectory to be corrected of the target aircraft model; A data processing unit, used for inputting the flight parameters extracted from the flight trajectory to be corrected into a pre-built parameter correction model to obtain correction parameters output by the parameter correction model; A result generating unit, used for performing reverse interpolation on the flight trajectory to be corrected based on the correction parameters to obtain a parameter correction result; The parameter correction model is obtained by training based on a convolutional neural network using labeled flight segments and a pre-stored initial parameter interpolation table; the labeled flight segments are obtained after data processing of real flight data samples.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images