Air refueling head wave modeling method based on neural network

Abstract

The invention belongs to the technical field of aircraft air refueling design, and discloses an air refueling head wave modeling method based on a neural network, and the method comprises the steps: obtaining the coordinates of a refueling machine taper sleeve relative to the tip of a fuel receiver taper pipe, and obtaining the coordinates of a refueling machine refueling pod relative to the tip of the fuel receiver taper pipe; adding refueling docking sort marks to the coordinates to make a data set, sorting the data set according to refueling docking sorts, calculating to obtain a closed external force of the taper sleeve under the refueling machine coordinate system, filtering data in the data set of each refueling docking sort, and subtracting a rope pulling force and a damping force borne by the taper sleeve to obtain a head wave field force borne by the taper sleeve; the coordinates of the oiling machine taper sleeve relative to the tip end of the oil receiver taper pipe serve as input, the head wave field force borne by the taper sleeve serves as output, the input is input into the deep neural network for training, and the deep neural network obtained through training is output to draw a head wave field cloud picture. According to the method, the stress condition of the taper sleeve in the oil receiving head wave field can be obtained, and then air refueling head wave modeling is completed.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of aircraft in-flight refueling design, and relates to an aircraft in-flight refueling head wave modeling method adopting a model fitting technology, in particular to an in-flight refueling head wave modeling method based on a neural network. BACKGROUND

[0002] Conventional in-flight refueling head wave modeling is completed based on wind tunnel tests, for example, a double-aircraft in-flight refueling docking wind tunnel free flight simulation device and method is disclosed in Chinese invention patent CN115541168A, the method comprises: the application is suitable for the technical field of wind tunnel tests, and provides a double-aircraft in-flight refueling docking wind tunnel free flight simulation device and method, which comprises a tanker model, a receiver model and a jet stream simulator; the jet stream simulator is fixed below the tanker model, the tanker model is installed in the wind tunnel test section through a support device fixed on the upper wall of the wind tunnel; the receiver model is a free flight model, and the receiver model is connected to a high-pressure gas pipe; the high-pressure gas pipe serves as the power source of the receiver model; the receiver model is arranged in the wake field of the tanker model. The method relies on wind tunnel test equipment, but the amount of test data obtained is not enough due to the shortage of wind tunnel resources and huge cost. SUMMARY

[0003] The purpose of the application: in order to solve the above problems, the application provides an in-flight refueling head wave modeling method based on a neural network, which processes the collected cone sleeve position as a data set, combines the dynamics model of the cone sleeve, obtains the force condition of the cone sleeve in the receiver head wave field, and then completes the in-flight refueling head wave modeling.

[0004] The technical scheme of the application: An in-flight refueling head wave modeling method based on a neural network comprises the following steps: S1, obtaining the coordinates of the tanker cone sleeve relative to the tip of the receiver cone pipe and the coordinates of the tanker refueling pod relative to the tip of the receiver cone pipe; S2, adding the coordinates obtained in S1 to the data set marked by the refueling docking frame number; S3, sorting the data set according to the refueling docking frame number, and calculating the cone sleeve force in the tanker coordinate system; S4, filtering the data in the data set of each refueling docking frame number respectively, and subtracting the rope tension and damping force suffered by the cone sleeve to obtain the head wave field force suffered by the cone sleeve; S5, taking the coordinates of the tanker cone sleeve relative to the tip of the receiver cone pipe as input and the head wave field force suffered by the cone sleeve as output, and inputting a deep neural network for training; S6, outputting the deep neural network obtained by training to draw a head wave field cloud chart.

[0005] Further, the coordinates in S1 are obtained by the aerial refueling camera.

[0006] Further, in S1, key points with known relative three-dimensional relationship are selected on the refueling machine or the cone sleeve, the pose relationship of the refueling machine or the cone sleeve relative to the camera is solved through PNP, and then the coordinates of the refueling machine cone tip relative to the refueling machine refueling pod relative to the refueling machine cone tip are solved according to the pose relationship of the camera relative to the cone tip.

[0007] Further, in S3, the resultant external force of the cone sleeve in the refueling machine coordinate system is obtained by twice derivation and multiplication by the mass of the cone sleeve.

[0008] Further, the filtering method in S4 is a smoothing filter.

[0009] Further, after S5 and before S6, the trained deep neural network is tested, and after the test is passed, S6 is performed.

[0010] Further, in S6, within the defined coordinate range, the coordinate input network is selected at intervals of 0.05m to obtain the network output of the deep neural network.

[0011] Further, the network output of the deep neural network is arranged in the form of a three-dimensional matrix to obtain a three-dimensional interpolation table, and a head wave field cloud chart is drawn according to the three-dimensional interpolation table.

[0012] Further, the defined coordinate range is: x:-3.5-0m, y:-2-3m, z:-2-2m.

[0013] Advantages of the present application: The present application proposes a neural network-based aerial refueling head wave modeling method for the aerial refueling scene. The processed cone sleeve position collected during the aerial refueling process is used as a data set, combined with the dynamics model of the cone sleeve, to obtain the force of the cone sleeve in the head wave field of the receiver, and then the aerial refueling head wave modeling is completed.

[0014] On the one hand, the present application is independent of the wind tunnel equipment compared with the wind tunnel test modeling method; on the other hand, the model accuracy is higher and closer to the actual head wave state compared with the CFD calculation method. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 A flowchart for the aerial refueling head wave modeling neural network of the present application is generated.

[0016] Figure 2 A flowchart for the neural network-based aerial refueling head wave modeling of the present application is generated. DETAILED DESCRIPTION

[0017] This part is the embodiment of the present application, used to explain and illustrate the technical solutions of the present application. In the case of no conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0018] In the description of the present application, it should be understood that the terms "center", "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship as given in the drawings, which are only for the purpose of facilitating the description of the present application and simplifying the description, and are not intended to indicate or imply that the device referred to or the case must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, the terms "first", "second" and the like are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implying the number of the technical features indicated. Therefore, the features defined with "first", "second" and the like can explicitly or implicitly include more features. In the description of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more.

[0019] In the description of the present application, it should be noted that unless otherwise specified and limited, the terms "mounting", "connection", "connection" should be interpreted broadly, for example, it can be fixed connection, or detachable connection or integrated connection; it can be mechanical connection, or point connection; it can be direct connection, or indirect connection through intermediate medium, or internal communication of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0020] Embodiment 1: A neural network-based air refueling head wave modeling method, comprising the following steps: S1, obtaining the coordinates of the refueling machine cone relative to the tip of the cone of the receiver, and obtaining the coordinates of the refueling machine refueling pod relative to the tip of the cone of the receiver; S2, adding the coordinates obtained in S1 to the data set of the refueling docking frame; S3, sorting the data set according to the refueling docking frame, and calculating the external force received by the cone in the refueling machine coordinate system; S4, filtering the data in the data set of each refueling docking frame, and subtracting the rope tension and damping force received by the cone, to obtain the head wave field force received by the cone; S5, taking the coordinates of the refueling machine cone relative to the tip of the cone of the receiver as input, and taking the head wave field force received by the cone as output, and inputting a deep neural network for training; S6, outputting the deep neural network obtained by training to draw a head wave field cloud chart.

[0021] The coordinates in S1 are obtained by an aerial refueling camera.

[0022] In S1, key points with known relative three-dimensional relationship are selected on the refueling machine or the cone sleeve, the pose relationship of the refueling machine or the cone sleeve relative to the camera is solved by PNP, and then the coordinates of the refueling machine cone tip relative to the refueling machine refueling pod relative to the refueling machine cone tip are solved according to the pose relationship of the camera relative to the cone tip.

[0023] In S3, the resultant force of the cone sleeve in the refueling machine coordinate system is obtained by twice derivation and multiplication of the mass of the cone sleeve.

[0024] The filtering method in S4 is a smoothing filter.

[0025] After S5 and before S6, the trained deep neural network is tested, and after the test is passed, S6 is performed.

[0026] In S6, within the limited coordinate range, the coordinates are selected as network inputs at intervals of 0.05m, and the network output of the deep neural network is obtained.

[0027] The network output of the deep neural network is arranged in the form of a three-dimensional matrix to obtain a three-dimensional interpolation table, and a head wave field cloud chart is drawn according to the three-dimensional interpolation table.

[0028] The limited coordinate range is: x:-3.5-0m, y:-2-3m, z:-2-2m.

[0029] Embodiment 2: A visual-based aerial refueling docking state recognition method, comprising: The coordinates of the cone sleeve relative to the refueling machine cone tip and the refueling machine refueling pod relative to the refueling machine cone tip are obtained by an aerial refueling camera / difference; The coordinates of the cone sleeve relative to the refueling machine refueling pod are calculated according to the cone sleeve relative to the refueling machine cone tip and the refueling machine refueling pod relative to the refueling machine cone; The data set is sorted according to each docking, and the force of the cone sleeve in the refueling machine coordinate system is obtained by twice derivation and multiplication of the mass of the cone sleeve; The data of the data set is smoothed and filtered, and the rope tension and damping force acting on the cone sleeve are subtracted to obtain the head wave field force acting on the cone sleeve; The position of the cone sleeve relative to the refueling machine cone tip is input, and the head wave field force acting on the cone sleeve is output, and the deep neural network is input for training; The position of the cone sleeve relative to the refueling machine cone tip is used as network input, and the head wave field force acting on the cone sleeve is compared with the network output to confirm whether the result is correct; The coordinates in the range of x:-3.5~0m, y:-2~3m, z-2~2m are selected as network inputs at intervals of 0.05m to obtain the network output. The network output is arranged into a three-dimensional matrix form to obtain a three-dimensional interpolation table; A head wave field cloud chart is drawn according to the three-dimensional interpolation table.

[0030] Key points with known relative three-dimensional relationships are selected on the refueling machine / cone sleeve, and a pose relationship of the refueling machine / cone sleeve relative to the camera is solved through PNP, and then a coordinate of the cone pipe tip of the receiver relative to the refueling machine refueling pod relative to the cone pipe tip of the receiver is solved according to the pose relationship of the camera relative to the cone pipe tip.

[0031] The selection criteria in x:-3.5~0m, y:-2~3m, z-2~2m are that the pilot manually docking trajectories are counted to ensure that the interpolation interval includes all successful docking trajectories.

[0032] In this example, the head wave modeling of aerial refueling based on a neural network is completed. The coordinates of the cone sleeve relative to the cone pipe tip of the receiver and the refueling machine refueling pod relative to the cone pipe tip of the receiver are obtained through aerial refueling cameras; the coordinates of the cone sleeve relative to the refueling machine refueling pod are calculated according to the cone sleeve relative to the cone pipe tip of the receiver and the refueling machine refueling pod relative to the cone pipe tip of the receiver; the first 11 docking sequences are sorted, the force of the cone sleeve in the refueling machine coordinate system is obtained by twice differentiation and multiplication of the cone sleeve mass; the data is filtered, and the rope tension and damping force borne by the cone sleeve are subtracted to obtain the head wave field force borne by the cone sleeve; the position of the cone sleeve relative to the cone pipe tip of the receiver is taken as the input, and the head wave field force borne by the cone sleeve is taken as the output, which is input into a deep neural network for training; the position of the cone sleeve relative to the cone pipe tip of the receiver for testing is taken as the network input, and the head wave field force borne by the cone sleeve and the network output are compared, and the proportion of state points with a force error greater than 5% of the statistical force interval range is less than 5%, and the model is considered reliable; the coordinates in the range of x:-3.5~0m, y:-2~3m, z-2~2m are selected as coordinate inputs into the network at intervals of 0.05m to obtain network outputs; the network outputs are arranged into a three-dimensional matrix form to obtain a three-dimensional interpolation table; and a head wave field cloud chart is drawn according to the three-dimensional interpolation table.

[0033] The above is only a specific embodiment of the present application, which is described in detail, and the part not described in detail is a conventional technology. However, the protection scope of the present application is not limited to this, any changes or replacements within the technical range disclosed by the present application can be easily thought by those skilled in the art, which should be covered within the protection scope of the present application. The protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for modeling the nosewave of aerial refueling based on neural networks, characterized in that, Includes the following steps: S1, obtain the coordinates of the refueling tanker cone sleeve relative to the tip of the receiver tanker cone, and obtain the coordinates of the refueling tanker refueling pod relative to the tip of the receiver tanker cone; S2, add refueling docking flight markers to the coordinates obtained in S1 to create a dataset; S3, sort the dataset by refueling docking order, and calculate the resultant external force on the cone sleeve in the refueling machine coordinate system; S4, filter the data in the dataset of each refueling docking flight, and subtract the rope tension and damping force on the cone sleeve to obtain the head wave field force on the cone sleeve; S5 takes the coordinates of the refueling tanker cone sleeve relative to the tip of the receiving tanker cone as input and the head wave field force on the cone sleeve as output, and inputs them into a deep neural network for training. S6, plots the output of the trained deep neural network as a head wave field cloud map.

2. The aerial refueling headwave modeling method based on neural networks according to claim 1, characterized in that, The coordinates in S1 were obtained using an aerial refueling camera.

3. The aerial refueling headwave modeling method based on neural networks according to claim 1, characterized in that, In S1, key points with known relative three-dimensional relationships are selected on the refueling aircraft or cone sleeve. The pose relationship between the refueling aircraft or cone sleeve and the camera is solved by PNP. Then, based on the pose relationship between the camera and the cone tip, the coordinates of the receiver aircraft cone tip and the refueling pod of the refueling aircraft relative to the receiver aircraft cone tip are solved.

4. The aerial refueling headwave modeling method based on neural networks according to claim 1, characterized in that, In S3, the net external force of the conical sleeve in the fuel tanker coordinate system is obtained by taking the second derivative and multiplying by the mass of the conical sleeve.

5. The aerial refueling headwave modeling method based on neural networks according to claim 1, characterized in that, The filtering method in S4 is smoothing filtering.

6. The aerial refueling headwave modeling method based on neural networks according to claim 1, characterized in that, After S5 and before S6, the trained deep neural network is tested. S6 is performed only after the test is passed.

7. The aerial refueling headwave modeling method based on neural networks according to claim 1, characterized in that, In S6, within a defined coordinate range, coordinates are selected and input into the network at 0.05m intervals to obtain the network output of the deep neural network.

8. The aerial refueling headwave modeling method based on neural networks according to claim 7, characterized in that, The outputs of the deep neural network are arranged into a three-dimensional matrix to obtain a three-dimensional interpolation table, and the head wave field cloud map is drawn based on the three-dimensional interpolation table.

9. The aerial refueling headwave modeling method based on neural networks according to claim 7, characterized in that, The defined coordinate range is: x: -3.5 to 0m, y: -2 to 3m, z: -2 to 2m.

Images