A method of simulating the motion of an aircraft store during release
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- AVIC XIAN AIRCRAFT IND GRP CO LTD
- Filing Date
- 2023-04-21
- Publication Date
- 2026-06-09
Smart Images
Figure CN116522770B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method for simulating the motion of objects, specifically, to a method for simulating the motion of aircraft external stores being deployed and separated. Background Technology
[0002] The separation process of external stores not only determines whether the stores can achieve their intended mission functions but also has a significant impact on the flight safety of the aircraft. During the release and separation of external stores, the aerodynamic forces of the flow field acting on them are a crucial factor determining their separation motion. Because the external stores and the aircraft are close together in the initial separation stage, their flow fields strongly interfere with each other. The intensity of this interference is related to their spatial position, making the release and separation motion of the external stores a complex coupled process. For a specific external store, the factors affecting its separation process are mainly the flight conditions of the aircraft, such as Mach number, altitude, angle of attack, and sideslip angle. Currently published technical literature primarily employs three methods to study the impact of release conditions on the separation process: numerical simulation, wind tunnel testing, and flight testing.
[0003] Flight tests and wind tunnel tests can obtain relatively accurate data on the separation motion of external attachments. However, in the safety analysis of external attachment deployment and separation, it is necessary to analyze the motion of as many state points as possible within the deployment envelope to determine the safety of the separation motion. If all studies on these deployment states are conducted using flight tests and wind tunnel tests, there are drawbacks such as long research cycles and high research costs. Numerical simulation methods, due to their advantages of short research cycles and relatively low costs, are currently widely used methods for studying deployment and separation motion. The commonly used numerical simulation method is the quasi-steady method based on computational fluid dynamics. This method assumes that the external attachment is in a steady flow field during separation, calculates the aerodynamic forces acting on the external attachment in the steady flow field, then solves the six-degree-of-freedom equations of motion of the external attachment, updates the spatial position of the external attachment at the next moment based on the calculation results of the equations of motion, updates the corresponding flow field of the external attachment, and recalculates the aerodynamic forces of the external attachment. The motion process of the external attachment during the separation time is calculated through this quasi-steady method. This method can calculate the separation trajectory and attitude of the external attachment relatively accurately, but the calculation is very time-consuming. Calculating and analyzing every state point within the aircraft's flight envelope is prohibitively time-consuming. Therefore, a simulation method capable of rapidly calculating the separation motion process based on the deployment status is needed. While neural networks are widely used in data fitting and nonlinear modeling, publicly available technical literature indicates a lack of application in external payload deployment and separation motion simulation. Applying neural networks to external payload separation motion simulation could potentially improve computational efficiency and reduce the research costs associated with deployment and separation safety analysis. Summary of the Invention
[0004] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method for simulating the release and separation motion of external attachments on an aircraft.
[0005] To achieve the above objectives, the present invention adopts the following technical solution:
[0006] A method for simulating the release and separation motion of external stores on an aircraft, comprising the following steps:
[0007] Step 1: Select several sets of deployment states within the aircraft's flight envelope to form a deployment state dataset;
[0008] Step 2: Set the initial state of the external object deployment, discretize the separation motion time of the external object into several sequence points to form a separation motion time dataset, obtain the motion trajectory and attitude of the external object at each time sequence point during the separation process, and form a motion dataset with the six degrees of freedom displacement of the external object's center of gravity.
[0009] Step 3: Merge the deployment state dataset from Step 1) and the separated motion time dataset from Step 2) into the input dataset of the neural network, and use the motion dataset from Step 2) as the output dataset of the neural network.
[0010] Step 4: Establish a single hidden layer feedforward neural network model, and use the input and output datasets from step 3) as the training data for the single hidden layer feedforward neural network model;
[0011] Step 5: Iteratively optimize the single hidden layer feedforward neural network model from Step 4). By optimizing the model's weights and biases, reduce the root mean square error of the model's fit until the number of iterations or the root mean square error of the fit is less than 10. -6 ;
[0012] Step 6: The single hidden layer feedforward neural network model optimized in step 5) is used as the external object deployment and separation motion model;
[0013] Step 7: Merge the deployment state dataset and the separation motion time dataset into an input dataset, input it into the single hidden layer feedforward neural network model obtained in step 6), calculate the separation motion trajectory and attitude of the external object, and realize the simulation of the motion of the external object separation process.
[0014] Preferably, the deployment status dataset in step 1 includes the aircraft's Mach number, altitude, angle of attack, and sideslip angle.
[0015] Preferably, the deployment status dataset selected in step 1 should be evenly distributed within the flight envelope of the carrier aircraft.
[0016] Preferably, in step 2, quasi-steady computational fluid dynamics methods or controllable trajectory force measurement experiments are used to calculate and obtain the motion trajectory and attitude of the external attachment at each time series point during the separation process.
[0017] Preferably, the input dataset formed by merging the deployment state dataset and the separated motion time dataset in step 3 is as shown in formula (1):
[0018]
[0019] In the formula: Ma, H, α, and β are the aircraft Mach number, altitude, angle of attack, and sideslip angle corresponding to the deployment state dataset, respectively. The subscript m represents the deployment state number, t is the separated motion time dataset, the subscript n represents the time sequence number, and the subscript l represents the product of the total number of deployment states and the total number of time sequence numbers. The output dataset is in the form of formula (2):
[0020]
[0021] In the formula: x, y, z, Φ, Θ, Ψ are the position and attitude angle of the external object during the separation motion, respectively; the subscript m represents the release state number; the subscript n represents the time sequence number; and the subscript l represents the product of the total number of release states and the total number of time sequence numbers.
[0022] Preferably, when establishing a single hidden layer feedforward neural network model in step 4, the number of nodes in the input layer of the neural network is 5, the hidden layer is a single layer with 12 to 16 nodes, and the number of nodes in the output layer is 6. The row vector of X in formula (1) is used as the input of the neural network, and the row vector Y in formula (2) is used as the output.
[0023] This simulation method learns the mapping relationship between deployment state and separation motion through a feedforward neural network. The trained model can quickly calculate the separation trajectory and attitude of the external attachment based on the deployment state. Compared with current quasi-steady calculation methods based on computational fluid dynamics, the simulation method and motion model of this invention have the advantages of fast calculation speed and high fitting accuracy. Attached Figure Description
[0024] Figure 1 This is a schematic diagram of the separation motion simulation method. Figure 2 The method calculates the separation motion trajectory and compares it with the actual value. In the figure, the solid line represents the actual value and the scatter points represent the calculated value of the model. Figure 3 To compare the model's predicted separation motion posture with the actual values, the solid line in the figure represents the actual values, and the scatter points represent the model's calculated values. Detailed Implementation
[0025] According to the present invention, an aircraft external stores release and separation motion simulation method is provided, the simulation process is as follows: Figure 1 As shown. The steps include:
[0026] Step 1: Uniformly select 120 sets of release status datasets, consisting of Mach number, altitude, angle of attack, sideslip angle, etc., within the flight envelope of the carrier aircraft;
[0027] Step 2: Set the initial position and attitude of the external attachment to 0. Simulate the separation motion process within 1 second after deployment. Discretize the separation motion time evenly into 20 time steps to form a separation motion time dataset. Then, use a quasi-steady method based on computational fluid dynamics to calculate the motion trajectory and attitude of the external attachment at each time series point during the separation process. The 6-DOF displacement of the external attachment's center of gravity forms a motion dataset.
[0028] Step 3: Organize the deployment status dataset, the separated motion time dataset, and the motion dataset into the forms shown in formulas (1) and (2):
[0029]
[0030]
[0031] Step 4: Establish a single hidden layer feedforward neural network model. Use the row vectors of dataset X as the training input of the neural network and the row vectors of dataset Y as the training output of the neural network. The hidden layer has 15 nodes, the input layer has 5 nodes, and the output layer has 6 nodes.
[0032] Step 5: Iteratively optimize the single hidden layer feedforward neural network model from Step 4. By optimizing the model's weights and biases, reduce the root mean square error of the model's fit until the number of iterations reaches 200 or the root mean square error of the fit is less than 10. -6 ;
[0033] Step 6: The single hidden layer feedforward neural network model optimized in Step 5 is used as the external object deployment and separation motion model;
[0034] Step 7: Randomly select a deployment state and use the fitted model to calculate the separation process of the attached object. The calculated trajectory of the attached object's separation motion is as follows: Figure 2 As shown, the calculated separation attitude is as follows Figure 3 As shown. Figure 2 , Figure 3 The solid line represents the actual values, and the scatter points represent the calculated values from the model. It can be seen that the model has a fairly high fitting accuracy. The model takes less than 1 second to calculate the separation motion for a single deployment state.
[0035] The specific embodiments of the present invention have been described above. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention. These improvements and modifications should also be within the scope of protection of the present invention.
Claims
1. A method for simulating the release and separation motion of external stores on an aircraft, characterized in that... Includes the following: 1) Select several sets of deployment states within the aircraft's flight envelope to form a deployment state dataset; 2) Set the initial state of the external object deployment, discretize the separation motion time of the external object into several sequence points to form a separation motion time dataset, obtain the motion trajectory and attitude of the external object at each time sequence point during the separation process, and form a motion dataset with the six degrees of freedom displacement of the external object's center of gravity. 3) Merge the deployment state dataset from step 1) and the separated motion time dataset from step 2) into the input dataset of the neural network, and use the motion dataset from step 2) as the output dataset of the neural network. 4) Establish a single hidden layer feedforward neural network model, and use the input and output datasets from step 3) as the training data for the single hidden layer feedforward neural network model. 5) Iteratively optimize the single hidden layer feedforward neural network model from step 4). By optimizing the model's weights and biases, reduce the root mean square error of the model's fit until the number of iterations or the root mean square error of the fit is less than 10. -6 ; 6) The single hidden layer feedforward neural network model optimized in step 5) is used as the external object deployment and separation motion model; 7) Merge the deployment state dataset and the separation motion time dataset into an input dataset, input it into the single hidden layer feedforward neural network model obtained in step 6), calculate the separation motion trajectory and attitude of the external attachment, and realize the simulation of the motion of the external attachment separation process.
2. The method for simulating the release and separation motion of aircraft external stores according to claim 1, characterized in that, The deployment status dataset in step 1) includes the aircraft's Mach number, altitude, angle of attack, and sideslip angle.
3. The method for simulating the release and separation motion of aircraft external stores according to claim 1, characterized in that, The deployment status dataset selected in step 1) should be evenly distributed within the flight envelope of the carrier aircraft.
4. The method for simulating the release and separation motion of aircraft external stores according to claim 1, characterized in that, In step 2), quasi-steady computational fluid dynamics methods or controllable trajectory force measurement experiments are used to calculate and obtain the motion trajectory and attitude of the external attachment at each time series point during the separation process.
5. The method for simulating the release and separation motion of aircraft external stores according to claim 1, characterized in that, The input dataset formed by merging the deployment state dataset and the separated motion time dataset in step 3) is shown in formula (1): In the formula: Ma, H, α, and β are the aircraft Mach number, altitude, angle of attack, and sideslip angle corresponding to the deployment state dataset, respectively. The subscript m represents the deployment state number, t is the separated motion time dataset, the subscript n represents the time sequence number, and the subscript l represents the product of the total number of deployment states and the total number of time sequence numbers. The output dataset is in the form of formula (2): In the formula: x, y, z, Φ, Θ, Ψ are the position and attitude angle of the external object during the separation motion, respectively; the subscript m represents the release state number; the subscript n represents the time sequence number; and the subscript l represents the product of the total number of release states and the total number of time sequence numbers.
6. The method for simulating the release and separation motion of aircraft external stores according to claim 1, characterized in that, In step 4), when establishing a single hidden layer feedforward neural network model, the number of nodes in the input layer of the neural network is 5, the hidden layer is a single layer with 12 to 16 nodes, and the number of nodes in the output layer is 6. The row vector of X in formula (1) is used as the input of the neural network, and the row vector Y in formula (2) is used as the output.