A launch vehicle offload control method and system based on an LSTM-FCNN attack angle estimation

Abstract

The application discloses a launch vehicle load reduction control method and system based on an attack angle estimation of an LSTM-FCNN, relates to the technical field of launch vehicle control, and comprises the following steps: generating a data set according to flight state data of a rocket in a high-altitude wind area and measured wind speed data, and processing the data set by adopting linear normalization; training an LSTM-FCNN deep neural network by using the normalized data, and establishing a mapping relationship between rocket attitude response and wind speed; migrating the trained LSTM-FCNN deep neural network to actual rocket flight, obtaining an attack angle based on estimated high-altitude wind speed, and then adopting an attack angle feedback-based load reduction control method to reduce the aerodynamic load of the launch vehicle. The application improves the estimation accuracy of the attack angle under uncertain wind field conditions, reduces the aerodynamic load of the launch vehicle, and improves the adaptability of the launch vehicle in high-altitude wind area flight.

Description

Technical Field

[0001] This invention relates to the field of launch vehicle control technology, specifically to a launch vehicle load reduction control method and system based on LSTM-FCNN angle of attack estimation. Background Technology

[0002] When a launch vehicle flies through the dense atmosphere, it experiences significant aerodynamic loads from high-altitude winds. If these loads exceed the structural strength of the rocket, the rocket will be damaged, catastrophically impacting the mission. Reducing the impact of high-altitude winds on launch vehicles, minimizing their aerodynamic loads, and ensuring safe and reliable flight in windy areas are challenging problems in the launch vehicle field. If load reduction control techniques can be effectively utilized to reduce the aerodynamic loads of launch vehicles, the design requirements for structural strength can be appropriately lowered, reducing the rocket's structural mass, increasing payload capacity, and reducing launch costs.

[0003] Currently, control schemes for reducing aerodynamic loads fall into two categories: passive load reduction, also known as "ballistic wind correction," and active load reduction. Passive load reduction utilizes wind field data to design the optimal trajectory, thereby minimizing the expected aerodynamic loads on the launch vehicle before launch. The compensation effect of passive load reduction depends primarily on the accuracy of the model and the accuracy of the pre-launch assessment of wind field patterns. Active load reduction, on the other hand, introduces angle-of-attack information into the control system, causing the launch vehicle to tend to fly against the airflow, thus reducing the rocket's angle of attack and sideslip angle during flight, and lowering the aerodynamic loads on the rocket. The key to active load reduction lies in the measurement of angle-of-attack information, which is generally divided into two methods: direct measurement and indirect measurement. Direct measurement obtains the flight angle of attack through angle-of-attack sensors, but considering measurement accuracy and installation and usage issues, it has not yet been incorporated into real-time control on launch vehicles. Indirect measurement uses information from accelerometers and extended state observers for load reduction control. However, directly introducing accelerometer signals into the control loop alters the control system's structure, thus affecting its stability. Extended state observers cannot distinguish between structural and wind disturbances, providing relatively accurate angle-of-attack estimations only when wind disturbances are dominant. Therefore, developing a high-precision angle-of-attack estimation method that comprehensively utilizes measurement information and relevant overall parameters is crucial for active load reduction control.

[0004] Chinese patent document CN117146660A discloses a rocket load reduction control method based on multi-source sensors, relating to the aerospace field. The method includes: considering multiple rocket parameters, designing a rocket attitude controller, and establishing a rocket load reduction control mission model; obtaining the rocket's normal acceleration using a load reduction accelerometer; the rocket's normal acceleration includes the normal acceleration caused by aerodynamic forces and the normal acceleration caused by engine thrust; during rocket flight, using an engine chamber pressure sensor to measure the chamber pressure in real time to determine the real-time engine thrust; determining the normal acceleration caused by aerodynamic forces based on the rocket's normal acceleration and engine thrust; adding feedback on the normal acceleration caused by aerodynamic forces to the rocket attitude controller; and determining the reduced total airflow angle of attack based on the rocket load reduction control mission model, thus achieving precise load reduction control of the rocket during the ascent phase. While this method can determine the normal acceleration caused by aerodynamic forces based on the rocket's normal acceleration and engine thrust, it relies on the measurement accuracy of the chamber pressure sensor and cannot eliminate the influence of structural interference on the normal acceleration.

[0005] Chinese patent document CN115729096A discloses a method and system for reducing the load on a sounding rocket. It employs an angle-of-attack sensor to measure the wind angle of attack, incorporating this real-time angle of attack into the sounding rocket's control system loop. By controlling engine oscillation, it compensates for wind interference, minimizing the load on the sounding rocket in high-wind areas, reducing wind-induced aerodynamic torque, and improving the launch adaptability of the sounding rocket. While this method can measure the angle of attack using an angle-of-attack sensor and then control the rocket's load reduction, current rockets cannot directly install angle-of-attack sensors to measure the angle of attack due to sensor installation limitations. Summary of the Invention

[0006] The purpose of this invention is to propose a launch vehicle load reduction control method and system based on LSTM-FCNN angle of attack estimation, which improves the estimation accuracy of angle of attack under uncertain wind conditions, reduces the aerodynamic load of the launch vehicle, and enhances the adaptability of the launch vehicle in high-altitude wind regions.

[0007] According to a first aspect of the present disclosure, a launch vehicle load reduction control method based on LSTM-FCNN angle of attack estimation is provided, comprising the following steps:

[0008] A dataset was generated based on the rocket's flight status data in high-altitude wind regions and the measured wind speed data, and then linear normalization was used to process the dataset.

[0009] The LSTM-FCNN deep neural network was trained using normalized data to establish the mapping relationship between rocket attitude response and wind speed.

[0010] The trained LSTM-FCNN deep neural network is transferred to the actual flight of the rocket. The angle of attack is obtained based on the estimated high-altitude wind speed. Then, a load reduction control method based on angle of attack feedback is adopted to reduce the aerodynamic load of the launch vehicle.

[0011] According to a second aspect of the present disclosure, a launch vehicle load reduction control system based on LSTM-FCNN angle of attack estimation is provided, comprising:

[0012] The data acquisition module generates a dataset based on the rocket's flight status data in the high-altitude wind region and the measured wind speed data, and processes the dataset using linear normalization.

[0013] The mapping relationship establishment module uses normalized data to train an LSTM-FCNN deep neural network to establish the mapping relationship between rocket attitude response and wind speed;

[0014] The load reduction control module transfers the trained LSTM-FCNN deep neural network to the actual flight of the rocket. Based on the estimated high-altitude wind speed, the angle of attack is obtained, and then a load reduction control method based on angle of attack feedback is adopted to reduce the aerodynamic load of the launch vehicle.

[0015] According to a third aspect of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and running on the memory, wherein the processor executes the program to implement the aforementioned method for launch vehicle load reduction control based on LSTM-FCNN angle of attack estimation.

[0016] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the aforementioned method for launch vehicle load reduction control based on LSTM-FCNN angle of attack estimation.

[0017] Compared with the prior art, the above technical solutions adopted in this invention have the following advantages: (1) In view of the limited accuracy of the launch vehicle angle of attack estimation, this invention uses LSTM-FCNN deep neural network to establish the mapping relationship between the rocket attitude response and wind speed, and estimates the angle of attack by estimating the wind speed, thereby effectively improving the accuracy of the angle of attack estimation for wind field uncertainty.

[0018] (2) In view of the excessive aerodynamic load during the flight of the launch vehicle, the present invention obtains the angle of attack information and introduces it into the control system, which can effectively reduce the aerodynamic load of the launch vehicle. Attached Figure Description

[0019] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an undue limitation of this application.

[0020] Figure 1 This is a block diagram illustrating the principle of a launch vehicle load reduction control method based on LSTM-FCNN angle of attack estimation. Specific implementation methods

[0021] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0022] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0023] 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.

[0024] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and systems according to various embodiments of this disclosure. It should be noted that each block in a flowchart or block diagram may represent a module, segment, or portion of code, which may include one or more executable instructions for implementing the logical functions specified in the various embodiments. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutively represented blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, may be implemented using a dedicated hardware-based system that performs the specified functions or operations, or using a combination of dedicated hardware and computer instructions.

[0025] Example 1:

[0026] like Figure 1 As shown, this embodiment provides a launch vehicle load reduction control method based on LSTM-FCNN angle of attack estimation, including the following steps:

[0027] Step 1: Generate a dataset based on the rocket's flight status data in the high-altitude wind region and the measured wind speed data, and process the dataset using linear normalization;

[0028] Specifically, to avoid large network prediction errors due to significant differences in magnitude, and to accelerate network training and improve prediction accuracy, this invention employs linear normalization as the data normalization method, with the following formula:

[0029]

[0030] Where x′ is the normalized data, the normalized data range is [0,1], and max(x) and min(x) are the maximum and minimum values ​​of the data, respectively.

[0031] Step 2: Train an LSTM-FCNN deep neural network using the normalized data to establish the mapping relationship between rocket attitude response and wind speed; specifically including:

[0032] Analyzing the motion of the launch vehicle's center of mass and its motion around the center of mass, the input quantities for the LSTM-FCNN deep neural network are determined as follows:

[0033]

[0034] Output quantity is

[0035]

[0036] Where t0 is the time starting point corresponding to each state variable in the sequence, t f As the end of time, Δψ is the pitch angle deviation, and Δψ is the yaw angle deviation. The apparent acceleration deviation in the y-direction. V represents the apparent acceleration deviation in the z-direction. fx Let V be the wind speed in the x-direction. fz Let H(t0) be the wind speed in the z-direction, and H(t0) be the rocket's altitude at time t0. f ) for t f The rocket's flight altitude at any given moment.

[0037] The long short-period neural network LSTM is used to process flight status data and wind speed data sequences, and the output of the long short-period neural network LSTM is extracted by combining it with a multi-layer fully connected neural network FCNN to obtain a more accurate mapping relationship that is closer to the actual situation. The activation function between each layer of the neural network is ReLU.

[0038] The LSTM-FCNN deep neural network is trained using normalized data. The Adam algorithm is employed for training, continuously adjusting network weights based on the loss function and evaluation metric until the network converges, completing the training. During training, the loss function is the mean squared error (MSE), and the evaluation metric is the root mean square error (RMSE), defined as follows:

[0039]

[0040] in y is the neural network prediction value. i is the true value, and n is the number of training samples.

[0041] The third step is to transfer the trained LSTM-FCNN deep neural network to the actual flight of the rocket, obtain the angle of attack based on the estimated high-altitude wind speed, and then adopt a load reduction control method based on angle of attack feedback to reduce the aerodynamic load of the launch vehicle.

[0042] The wind speed is estimated online using a trained LSTM-FCNN deep neural network, and the angle of attack is obtained based on the wind speed, specifically as follows:

[0043] Transferring rocket velocity from the launch system to the rocket system:

[0044]

[0045] in V is the coordinate transformation matrix from the rocket launch system to the rocket system. x V y V z Let V be the rocket's velocity relative to the ground in the three directions in the launch coordinate system. qxB V qyB V qzB Let be the rocket's air velocity in three directions within the rocket's coordinate system;

[0046] The rocket's angle of attack and sideslip angle are obtained based on its anti-air velocity within the rocket system, as follows:

[0047]

[0048] In the above formula, V qB Let be the rocket's air velocity in the rocket body coordinate system, and α1 and β1 be the total angle of attack and total sideslip angle, respectively. The specific expressions are as follows:

[0049]

[0050] Where α is the response angle of attack, α w β is the angle of attack of the airflow, and β is the sideslip angle of response. w It represents the sideslip angle of the airflow.

[0051] After estimating the angle of attack, the angle of attack information is incorporated into the control system. The control equations for the rocket's pitch channel are as follows:

[0052]

[0053] in This represents the attitude angle feedback gain coefficient. This represents the attitude angular velocity feedback gain coefficient. Here, Δα is the angle-of-attack feedback load reduction gain coefficient, and Δα is the response angle-of-attack deviation. This is the engine sway angle command.

[0054] When considering the effects of wind shear, the equation of motion for the rocket's pitch attitude can be expressed as:

[0055]

[0056] Where α wq The coefficient is the angle of attack of the airflow generated by the shear wind. The expression is as follows:

[0057]

[0058] in The moment of inertia of the pitch channel, Let q be the derivative of the normal aerodynamic coefficient with respect to α, and S be the dynamic pressure. M Let X be the reference cross-sectional area of ​​the rocket. d X is the distance from the theoretical tip of the arrow body to the aerodynamic center of pressure. z Let P be the theoretical distance from the tip of the rocket body to its center of mass, and X be the rocket thrust. R Let m be the distance from the engine's swing point to the theoretical tip of the rocket body. R The mass of the oscillating part of a single engine. l is the apparent longitudinal acceleration of the arrow body. R This is the distance from the engine's center of mass to the pivot axis.

[0059] Further combining equations (8) and (9), we obtain the total angle of attack deviation caused by the shear wind as follows:

[0060]

[0061] Equation (12) shows that after introducing the angle of attack information estimated by the neural network, by selecting an appropriate... The chosen value can reduce the total angle of attack, thereby reducing the aerodynamic load on the rocket body and improving the adaptability of the launch vehicle to fly in high-altitude wind regions.

[0062] This method can resist the influence of structural interference forces and measurement noise under uncertain wind field conditions, and improve the accuracy of angle of attack estimation. It does not require additional measurement equipment, is easy to implement, can effectively reduce the aerodynamic load of the rocket, and improve the adaptability of the launch vehicle to fly in high-altitude wind areas.

[0063] Example 2:

[0064] This embodiment provides a launch vehicle load reduction control system based on LSTM-FCNN angle of attack estimation, including:

[0065] The data acquisition module generates a dataset based on the rocket's flight status data in the high-altitude wind region and the measured wind speed data, and processes the dataset using linear normalization.

[0066] The mapping relationship establishment module uses normalized data to train an LSTM-FCNN deep neural network to establish the mapping relationship between rocket attitude response and wind speed;

[0067] The load reduction control module transfers the trained LSTM-FCNN deep neural network to the actual flight of the rocket. Based on the estimated high-altitude wind speed, the angle of attack is obtained, and then a load reduction control method based on angle of attack feedback is adopted to reduce the aerodynamic load of the launch vehicle.

[0068] Example 3:

[0069] An electronic device includes a memory, a processor, and a computer program stored in the memory and running thereon. When the processor executes the program, it implements the aforementioned method for launch vehicle load reduction control based on LSTM-FCNN angle-of-attack estimation, comprising:

[0070] A dataset was generated based on the rocket's flight status data in high-altitude wind regions and the measured wind speed data, and then linear normalization was used to process the dataset.

[0071] The LSTM-FCNN deep neural network was trained using normalized data to establish the mapping relationship between rocket attitude response and wind speed.

[0072] The trained LSTM-FCNN deep neural network is transferred to the actual flight of the rocket. The angle of attack is obtained based on the estimated high-altitude wind speed. Then, a load reduction control method based on angle of attack feedback is adopted to reduce the aerodynamic load of the launch vehicle.

[0073] Example 4:

[0074] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned method for launch vehicle load reduction control based on LSTM-FCNN angle of attack estimation, comprising:

[0075] A dataset was generated based on the rocket's flight status data in high-altitude wind regions and the measured wind speed data, and then linear normalization was used to process the dataset.

[0076] The LSTM-FCNN deep neural network was trained using normalized data to establish the mapping relationship between rocket attitude response and wind speed.

[0077] The trained LSTM-FCNN deep neural network is transferred to the actual flight of the rocket. The angle of attack is obtained based on the estimated high-altitude wind speed. Then, a load reduction control method based on angle of attack feedback is adopted to reduce the aerodynamic load of the launch vehicle.

[0078] Those skilled in the art will understand that the modules or steps described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, which can then be stored in a storage device for execution by a computer device. Alternatively, they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. This disclosure is not limited to any particular combination of hardware and software.

[0079] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0080] While the specific embodiments of this disclosure have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of this disclosure. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of this disclosure are still within the scope of protection of this disclosure.

Claims

1. A launch vehicle offload control method based on LSTM-FCNN attack angle estimation, characterized in that, The method comprises the following steps: A data set is generated according to flight state data of the rocket in a high-altitude wind area and measured wind speed data, and the data set is processed by linear normalization; An LSTM-FCNN deep neural network is trained by using the normalized data to establish a mapping relationship between a rocket attitude response and a wind speed; The trained LSTM-FCNN deep neural network is migrated to actual rocket flight, an angle of attack is obtained based on an estimated high-altitude wind speed, and a load reduction control method based on angle of attack feedback is adopted to reduce aerodynamic load of the launch vehicle; The input quantity of the LSTM-FCNN deep neural network is determined by analyzing mass center motion and motion around the mass center of the launch vehicle: The output quantity is: wherein is the time start point corresponding to each state quantity in the sequence, is the time end point, is the pitch angle deviation, is the yaw angle deviation, is the y-direction visual acceleration deviation, is the z-direction visual acceleration deviation, is the x-direction wind speed, is the z-direction wind speed, is is the rocket flight height at the moment, is is the rocket flight height at the moment; The wind speed is estimated online by using the trained LSTM-FCNN deep neural network, and the angle of attack is obtained based on the wind speed, specifically as follows: The rocket speed is converted from a launch system to a rocket system: wherein is the coordinate transformation matrix from the rocket launch system to the vehicle system, , , is the rocket velocity in the three directions of the launch coordinate system, , , is the rocket velocity in the three directions of the vehicle coordinate system; The rocket angle of attack and sideslip angle are obtained based on the air speed in the rocket system as follows: In the above formula is the airspeed of the rocket in the body coordinate system, , are the total attack angle and the total sideslip angle, respectively, and the specific expressions are as follows: wherein is the angle of attack, is the angle of attack of the airflow, is the angle of sideslip, is the angle of sideslip of the airflow; The angle of attack information is introduced into the control system, and the rocket pitch channel control equation is as follows: wherein is a feedback gain coefficient for the attitude angle, is a feedback gain coefficient for the attitude angular velocity, is a feedback gain coefficient for the angle of attack, is a response angle of attack deviation, is an engine swing angle command.

2. The control method of claim 1, wherein the LSTM-FCNN is configured to estimate the angle of attack based on the plurality of flight parameters. In the LSTM-FCNN deep neural network, a long-short period neural network LSTM is first used to process flight state data and wind speed data sequences, and a multi-layer fully connected neural network FCNN is combined to extract features of the output quantity of the long-short period neural network LSTM, so that a mapping relationship close to the actual situation is obtained.

3. The control method of claim 1, wherein the LSTM-FCNN is based on an angle of attack estimation. When the shear wind effect is considered, the rocket pitch attitude motion equation is represented as: wherein is the angle of attack of the airflow generated by the shear wind, the coefficient , The expression is as follows: in The moment of inertia of the pitch channel, For the normal aerodynamic coefficient pair The derivative of For dynamic pressure, The reference cross-sectional area of ​​the rocket. The distance from the theoretical tip of the arrow body to the aerodynamic center of pressure. Let be the theoretical distance from the tip of the arrow body to its center of mass. For rocket thrust, This is the distance from the engine's swing point to the theoretical tip of the rocket body. The mass of the oscillating part of a single engine. The longitudinal apparent acceleration of the arrow body. This is the distance from the engine's center of mass to the pivot axis.

4. The control method of claim 3, wherein the LSTM-FCNN attack angle estimation is based on a Long Short-Term Memory (LSTM) and a Fully Convolutional Neural Network (FCNN). The total angle of attack deviation caused by the shear wind is: The above formula shows that the total attack angle is reduced by selecting value, thereby reducing the aerodynamic load of the rocket body and improving the adaptability of the launch vehicle in the high-altitude wind area.

5. A launch vehicle offload control system based on LSTM-FCNN angle of attack estimation for implementing the method of any one of claims 1-4, characterized in that, The method comprises the following steps: A data acquisition module generates a data set according to flight state data of the rocket in a high-altitude wind area and measured wind speed data, and processes the data set by linear normalization; A mapping relationship establishment module trains an LSTM-FCNN deep neural network by using the normalized data to establish a mapping relationship between a rocket attitude response and a wind speed; A load reduction control module migrates the trained LSTM-FCNN deep neural network to actual rocket flight, obtains an angle of attack based on an estimated high-altitude wind speed, and adopts a load reduction control method based on angle of attack feedback to reduce aerodynamic load of the launch vehicle.

6. An electronic device comprising a memory, a processor, and a computer program stored on the memory to run on the processor, characterized in that, The processor executes the program to implement the launch vehicle load reduction control method based on LSTM-FCNN angle of attack estimation according to any one of claims 1-4.

7. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the launch vehicle load reduction control method based on LSTM-FCNN angle of attack estimation according to any one of claims 1-4.

Images