A method and device for suppressing muzzle vibration
By predicting muzzle vibration based on a time-series deep learning model and generating reverse suppression commands, the problem of muzzle vibration affecting firing accuracy during continuous firing of medium and large caliber gun barrels is solved, and rapid suppression and improvement of firing accuracy are achieved.
Patent Information
- Application Number
- CN202410343646.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-25
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2044-03-25
AI Technical Summary
When medium and large caliber gun barrels are firing in bursts, the muzzle will still be vibrating when the next shot is fired, which affects the firing accuracy.
A time-series-based deep learning model is used to predict the muzzle vibration curve and generate a reverse vibration suppression command. The data processing module and controller control the servo system to achieve the prediction and suppression of muzzle vibration.
While maintaining a constant rate of fire, the muzzle vibration is quickly suppressed to ensure that there is no muzzle vibration during the next shot, thereby improving firing accuracy.
Smart Images

Figure CN118258259B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method and device for suppressing muzzle vibration, belonging to the field of servo system control technology for medium and large caliber gun barrels. Background Technology
[0002] Medium and large caliber gun barrels are large in size and weight, and have high muzzle kinetic energy. The muzzle vibration amplitude caused by the instant the projectile leaves the muzzle is large and the decay time is long, which directly affects the firing rate and firing accuracy of the equipment.
[0003] The firing process of medium-to-large caliber weapons is as follows: the projectile is given an initial velocity, and after acquiring this velocity, it rapidly detaches from the weapon and reaches the target point. According to firing dynamics theory, because the projectile generates a very high velocity in a short period of time, the firing system will be subjected to a huge load impact during the firing process. The large-caliber barrel, under the influence of numerous external forces (such as the combined force of the barrel, recoil force, and projectile gravity) and its own gravity during firing, will experience axial and lateral vibrations, thereby reducing firing accuracy and affecting the firing rate. Existing vibration suppression strategies are based on PID controllers and aiming transmission devices in the servo system. The main steps are that during firing, the PID controller of the servo system applies large external damping energy to the barrel through the aiming transmission device, thereby accelerating the decay of the vibration period and reducing the amplitude. Although this method can effectively suppress the vibration state of the muzzle, the suppression time is relatively long. However, the time interval between adjacent shots in the current continuous firing is relatively short, so the vibration suppression time exceeds the time interval between adjacent shots. This results in the muzzle still being affected by the previous shot and in a state of vibration when the next shot is fired, thus affecting the firing accuracy. Therefore, this muzzle vibration suppression method is not suitable for high-rate-of-fire cannons. Summary of the Invention
[0004] The purpose of this invention is to provide a method and apparatus for suppressing muzzle vibration, in order to solve the problem of large muzzle disturbance and low firing accuracy caused by the muzzle still being in a vibrating state when firing the next shot during continuous firing.
[0005] To achieve the above objectives, the present invention includes:
[0006] A method for suppressing muzzle vibration according to the present invention includes the following steps:
[0007] Upon receiving a firing command, the fire control angle command, ammunition mass, and firing energy at the moment of firing are recorded. The firing angle, ammunition mass, and firing energy in the fire control angle command are then input into the trained prediction model to obtain the predicted muzzle vibration curve corresponding to the fire control angle command.
[0008] The muzzle vibration suppression command is determined based on the predicted muzzle vibration curve, and the controlled object is controlled according to the muzzle vibration suppression command in order to suppress the vibration caused by firing the current fire control angle command. The controlled object is the gun barrel.
[0009] The prediction model adopts a time series-based deep learning model; the dataset used for training the prediction model includes the ammunition mass of the cannon barrel, the firing angle under fixed-angle firing of the cannon barrel, the firing energy, and the corresponding muzzle vibration curve.
[0010] Furthermore, the muzzle vibration suppression command is a vibration curve with the same vibration period, the same amplitude, and the opposite amplitude direction as the predicted muzzle vibration curve.
[0011] Furthermore, the time series prediction model is a long short-term memory deep network model.
[0012] Furthermore, the expression for the error function used during the training of the prediction model is as follows:
[0013]
[0014] In the formula, MSE is the error function, y k For the kth actual value, Let K be the k-th predicted value, and K be the total number of predicted values.
[0015] Furthermore, the data in the dataset is standardized data.
[0016] Furthermore, the method also includes a pre-firing aiming phase, which is implemented as follows: real-time acquisition of the fire control angle command and the current angle position of the controlled object; obtaining a position command based on the current angle position of the controlled object and the fire control angle command; and controlling the controlled object based on the position command.
[0017] Furthermore, the method also includes a tracking and aiming stage after vibration suppression. The implementation steps of this stage are as follows: real-time acquisition of the fire control firing angle command of the controlled object and the current firing angle position of the controlled object; obtaining a position command based on the current firing angle position of the controlled object and the fire control firing angle command; and controlling the controlled object based on the position command.
[0018] Beneficial effects: The muzzle vibration suppression method of the present invention is a further improvement on the existing vibration suppression method. The present invention predicts the muzzle vibration curve based on the prediction model, and generates a muzzle vibration suppression command that can be reversed based on the muzzle vibration curve. Vibration suppression can be performed according to the muzzle vibration suppression command, so as to suppress the vibration caused by firing the current fire control angle command during continuous firing. This method can suppress most of the vibration energy, so that the muzzle vibration amplitude can be attenuated faster and the total muzzle vibration time can be shortened faster. Under the condition of keeping the firing rate of medium and large caliber guns constant, the muzzle is in a vibration-free state at the moment of the next firing during continuous firing, avoiding the initial disturbance and the problem of affecting the firing accuracy caused by muzzle vibration.
[0019] This invention discloses a muzzle vibration suppression device, comprising a data processing module and a controller connected in communication. The data processing module, upon receiving a firing command, calculates a predicted muzzle vibration curve based on the fire control firing angle command, ammunition mass, and firing energy data at the current firing moment, and generates a muzzle vibration suppression command based on the predicted muzzle vibration curve. The controller receives the muzzle vibration suppression command from the data processing module and controls the gun barrel according to the command to suppress vibrations caused by firing at the current fire control firing angle command. The data processing module generates the predicted muzzle vibration curve using a prediction model, which employs a time-series-based deep learning model. The dataset used for training the prediction model includes the ammunition mass of the gun barrel, the firing angle under fixed-angle firing, the firing energy, and the corresponding muzzle vibration curve.
[0020] Furthermore, the prediction model is a trained long short-term memory deep network model.
[0021] Furthermore, the muzzle vibration suppression command is a vibration curve with the same vibration period, the same amplitude, and the opposite amplitude direction as the predicted muzzle vibration curve.
[0022] Beneficial effects: The muzzle vibration suppression device of the present invention, through the data processing module and controller, can effectively realize a muzzle vibration suppression method, suppressing most of the vibration energy, thereby achieving faster attenuation of the muzzle vibration amplitude and faster reduction of the total muzzle vibration time; under the condition of maintaining a constant firing rate of medium and large caliber guns, the muzzle is in a vibration-free state at the moment of the next firing during continuous firing, avoiding the initial disturbance and impact on firing accuracy caused by muzzle vibration. Attached Figure Description
[0023] Figure 1 This is a schematic diagram of the muzzle vibration suppression device in an embodiment of the present invention;
[0024] Figure 2 This is a schematic diagram of the predictive model in an embodiment of the device of the present invention;
[0025] Figure 3 This is a schematic diagram of the muzzle vibration suppression control process in an embodiment of the present invention;
[0026] Figure 4 This is a typical muzzle vibration angular displacement curve diagram in the device embodiment of the present invention;
[0027] Figure 5 This is a muzzle vibration angular displacement curve obtained by using a deep learning model in an embodiment of the device of the present invention.
[0028] Figure 6 This is a muzzle vibration curve diagram after the action of the servo system controlled by the controller is coupled with the forced vibration of the muzzle in the device embodiment of the present invention. Detailed Implementation
[0029] The present invention will now be described in further detail with reference to the accompanying drawings.
[0030] The present invention provides a method and device for suppressing muzzle vibration. The suppression method mainly involves: upon receiving a firing command, recording the fire control angle command, ammunition mass, and firing energy at the firing moment; inputting the firing angle, ammunition mass, and firing energy from the fire control angle command into a trained prediction model to obtain a predicted muzzle vibration curve corresponding to the fire control angle command; forming a muzzle vibration suppression command based on the predicted muzzle vibration curve; and controlling the controlled object according to the muzzle vibration suppression command to suppress the vibration caused by firing the current fire control angle command. The controlled object refers to the barrel of the gun.
[0031] Example of a muzzle vibration suppression device:
[0032] like Figure 1 The muzzle vibration suppression device shown mainly includes a data processing module and a controller connected by communication. The device also includes a follow-up system and an aiming transmission device.
[0033] The data processing module includes a trained prediction model that, upon receiving a firing command, can obtain a predicted muzzle vibration curve based on the fire control angle command, ammunition mass, and firing energy at the currently recorded firing moment. It can then use the predicted muzzle vibration curve to generate a muzzle vibration suppression command, keeping the current fire control angle command unchanged.
[0034] Considering the need for advance prediction and early suppression, time-series based networks, after training, can well meet this function. Of course, other networks may also be able to solve this problem. Specifically, the prediction model adopts a time-series based deep learning model, and in this embodiment, the prediction model is a Long Short-Term Memory (LSTM) deep network model.
[0035] This Long Short-Term Memory (LSTM) deep network model architecture includes a forget gate, an input gate, and an output gate, such as... Figure 2 As shown, the forget gate processes the previous state information, the input gate controls the information to be input into the network, and the output gate determines the muzzle vibration curve after updating the state. The specific LSTM structure is as follows:
[0036]
[0037] Where: h t Output vector for the current time step; x t z is the input vector at the current time step. t i t f t c t o t h t These are, respectively, the input signal, the input gate, the forget gate, the state unit, the output gate, and the output signal; (W) z W i W f W o ), (U z U i U f U o (b) z b i b f b o The input signals z and z are respectively. t Input gate i t Forgotten Gate t Output gate o t The input weight matrix, recursive weight matrix, and bias matrix are given; σ is the sigmoid activation function, and g is the tanh function.
[0038] The Long Short-Term Memory (LSTM) deep network model is trained using the acquired dataset. The dataset includes ammunition mass of the caliber weapon, the firing angle of the caliber weapon at a fixed angle, the firing energy, and the corresponding muzzle vibration curve. Specifically, muzzle vibration curves of large-caliber guns under different ammunition masses, different firing energies, and different representative typical firing angles are acquired sequentially to construct a muzzle vibration input-output signal dataset. To facilitate subsequent model training, the data in the dataset is standardized. In this embodiment, the controlled object is a medium-to-large caliber new concept gun. The muzzle vibration suppression command in this embodiment is a vibration curve with the same vibration period, equal amplitude, and opposite amplitude direction as the predicted muzzle vibration curve.
[0039] The model is trained using data from the dataset. In this embodiment, the prediction model is configured with a predetermined number of LSTM hidden layers (e.g., 200 in this embodiment), one fully connected layer, and one regression layer. The loop dropout is set to 0.2; the maximum number of iterations is a predetermined number, e.g., 1000 iterations. The prediction model converges using the Sigmoid activation function f. x The model error MSE uses the actual value y. k and target value The mean squared error is used for calculation, and the model validation MAE uses the actual value y. k and target value The mean absolute error is used for calculation. After constructing and training the model, a prediction model is obtained. The sigmoid activation function f is used. x The calculation expression is:
[0040]
[0041] The expression for calculating the error function MSE used during the training of the prediction model is as follows:
[0042]
[0043] The calculation expression for the validation function MAE used during the training of the prediction model is as follows:
[0044]
[0045] In the formula, MSE is the error function, y k For the kth actual value, Let K be the k-th predicted value, K be the total number of predicted values, and MAE be the validation function.
[0046] The controller receives muzzle vibration suppression commands from the data processing module and controls the servo system accordingly. Specifically, the controller reads the real-time position of the gun, calculates the muzzle vibration suppression commands, and then controls the servo system to act on the aiming transmission device.
[0047] The servo system generates forces acting on the aiming transmission device based on the muzzle vibration suppression command from the controller. Specifically, the servo system mainly includes azimuth and elevation systems, which provide forces along the elevation and azimuth axes, respectively.
[0048] The aiming transmission device is used for mechanical connection with the controlled object, and is used to drive the controlled object to move according to the force of the servo system. In this embodiment, the aiming transmission device is an actuator, capable of driving the controlled object to move in two axial directions: elevation and yaw. In this embodiment, the controlled object is a large-caliber gun.
[0049] like Figure 3 As shown, the entire firing process of a caliber cannon includes a pre-firing tracking phase, a post-firing suppression phase, and a post-suppression tracking phase. y0 represents the pre-firing fire control command, y2 represents the suppressed fire control command (no fire control commands are accepted during suppression), e represents the error, and u represents the control quantity. The post-suppression tracking phase uses the same tracking method as the pre-firing phase. The process is as follows: The fire control angle command and the current angle position of the controlled object are acquired in real time. The control error is obtained based on the current angle position and the fire control angle command. This control error is then fed into the controller to generate a control quantity, which drives the servo system to achieve the tracking objective.
[0050] When the servo system receives the firing command, this device executes the following control process: records the fire control angle command r1, ammunition mass m1, and firing energy J1 at the moment of firing; inputs the fire control angle command r1, ammunition mass m1, and firing energy J1 at this moment into the trained prediction model; and outputs the predicted muzzle vibration curve y1.
[0051] At this time, the current fire control firing angle command r1 remains unchanged. Based on the predicted muzzle vibration curve y1, a muzzle vibration suppression command -y1 is generated. After the muzzle vibration suppression command -y1 is calculated by the controller, it is input to the follow-up system. The follow-up system generates forces on two axes: elevation and yaw. After this firing is completed, it drives the aiming transmission device to control the controlled object to quickly reduce and cancel the vibration energy of the elevation part.
[0052] The vibration period in the motion curve corresponding to the muzzle vibration suppression command -y1 is equal to the caliber vibration period of the predicted muzzle vibration curve, with equal amplitude and opposite direction.
[0053] After firing is completed, the servo system resumes real-time reception of fire control firing angle command r2 and position feedback y of the controlled object, and controls the controlled object to track the target in real time according to the control method of the pre-firing tracking and aiming stage.
[0054] As shown in Figure 4, a typical muzzle vibration angular displacement curve is obtained after prediction by the prediction model in this embodiment. Figure 5 The predicted values and the actual measured values basically coincide. After the predicted data is passed through the controller to control the servo system for vibration suppression, the characteristics after suppression can be observed as follows: Figure 6 As shown, the vibration is significantly attenuated to a level that does not affect shooting accuracy within the first cycle, and after suppression, it has virtually no impact on shooting accuracy.
[0055] In summary, this embodiment uses muzzle vibration angular displacement data during fixed-angle firing and employs end-to-end deep learning to learn the muzzle vibration curve based on time series. This allows for a more realistic and reasonable prediction of the muzzle vibration amplitude, period, and attenuation rate, resulting in a more accurate, reasonable, and efficient reverse enable for the servo system to suppress vibration energy. The muzzle vibration suppression method, based on deep learning prediction, can suppress vibration at the moment of initiation, eliminating a large portion of the vibration energy. This leads to faster attenuation of the muzzle vibration amplitude and a faster reduction in the total muzzle vibration time. While maintaining a constant firing rate for medium and large caliber guns, the muzzle is in a vibration-free state at the moment of the next firing during continuous fire, avoiding initial disturbances and impacts on firing accuracy caused by muzzle vibration.
[0056] Example of a muzzle vibration suppression method:
[0057] The muzzle vibration suppression method in this embodiment mainly includes the following steps:
[0058] Upon receiving a firing command, the fire control angle command, ammunition mass, and firing energy at the moment of firing are recorded. The fire control angle command, ammunition mass, and firing energy are then input into the trained prediction model to obtain the predicted muzzle vibration curve.
[0059] The muzzle vibration suppression command is determined based on the predicted muzzle vibration curve. The controlled object, the gun barrel, is then controlled according to this command to suppress the vibration caused by firing at the current fire control angle. The prediction model employs a time-series-based deep learning model. The dataset used for training the model includes the ammunition mass of the gun barrel, the firing angle under a fixed firing angle, the firing energy, and the corresponding muzzle vibration curve. Specific implementation steps and principles have been described in detail in the section on muzzle vibration suppression devices and will not be repeated here.
Claims
1. A method for suppressing muzzle vibration, characterized in that, Includes the following steps: Upon receiving a firing command, the fire control angle command, ammunition mass, and firing energy at the moment of firing are recorded. The firing angle, ammunition mass, and firing energy in the fire control angle command are then input into the trained prediction model to obtain the predicted muzzle vibration curve corresponding to the fire control angle command. The muzzle vibration suppression command is determined based on the predicted muzzle vibration curve. The controlled object is then controlled according to the muzzle vibration suppression command to suppress the vibration caused by firing the current fire control angle command. The controlled object is the gun barrel. The muzzle vibration suppression command is a vibration curve with the same vibration period, the same amplitude, and the opposite amplitude direction as the predicted muzzle vibration curve. The prediction model employs a long short-term memory deep network model; the dataset used during training of the prediction model includes the ammunition mass of the cannon barrel, the firing angle under fixed-angle firing of the cannon barrel, the firing energy, and the corresponding muzzle vibration curve; The method also includes a pre-firing aiming phase, which is implemented as follows: real-time acquisition of the fire control angle command and the current angle position of the controlled object; obtaining a position command based on the current angle position and the fire control angle command; and controlling the controlled object based on the position command.
2. The method for suppressing muzzle vibration according to claim 1, characterized in that, Controlling the controlled object according to the muzzle vibration suppression command includes: calculating the muzzle vibration suppression command and then controlling the servo system to act on the aiming transmission device. The aiming transmission device is used to mechanically connect with the controlled object and to drive the controlled object to move according to the force of the servo system.
3. The method for suppressing muzzle vibration according to claim 1, characterized in that, The prediction model is configured with a set number of LSTM hidden layers with neurons, one fully connected layer, and one regression layer.
4. The method for suppressing muzzle vibration according to claim 3, characterized in that, The expression for the error function used during prediction model training is as follows: In the formula, MSE is the error function. For the kth actual value, Let K be the k-th predicted value, and K be the total number of predicted values.
5. The method for suppressing muzzle vibration according to claim 1 or 3, characterized in that, The data in the dataset is standardized data.
6. The method for suppressing muzzle vibration according to claim 2, characterized in that, The servo system mainly includes a directional rotation system and a pitch system, which respectively provide forces along the pitch and directional rotation axes.
7. The method for suppressing muzzle vibration according to claim 1, characterized in that, The method also includes a tracking and aiming phase after vibration suppression. The implementation steps of this phase are as follows: real-time acquisition of the fire control firing angle command of the controlled object and the current firing angle position of the controlled object; obtaining a position command based on the current firing angle position of the controlled object and the fire control firing angle command; and controlling the controlled object based on the position command.
8. A muzzle vibration suppression device, characterized in that, The system includes a data processing module and a controller with communication connectivity. The data processing module, upon receiving a firing command, calculates a predicted muzzle vibration curve based on the fire control angle command, ammunition mass, and firing energy data at the current firing moment, and generates a muzzle vibration suppression command based on the predicted muzzle vibration curve. The controller receives the muzzle vibration suppression command from the data processing module and controls the barrel accordingly to suppress vibrations caused by firing at the current fire control angle command. The muzzle vibration suppression command is a vibration period and amplitude equal to the predicted muzzle vibration curve. The data processing module generates predicted muzzle vibration curves using a prediction model, which employs a trained long short-term memory deep network model. The dataset used during training of the prediction model includes the ammunition mass of the barrel, the firing angle under fixed-angle firing, the firing energy, and the corresponding muzzle vibration curves. The controller is also used to acquire the fire control firing angle command and the current firing angle position of the controlled object in real time during the pre-fire tracking phase, obtain the position command based on the current firing angle position and the fire control firing angle command, and control the controlled object according to the position command.
9. The muzzle vibration suppression device according to claim 8, characterized in that, The prediction model is configured with a set number of LSTM hidden layers with neurons, one fully connected layer, and one regression layer.
10. The muzzle vibration suppression device according to claim 8, characterized in that, The controller controls the barrel according to the muzzle vibration suppression command, including: the controller calculates the muzzle vibration suppression command and controls the follow-up system to act on the aiming transmission device, the aiming transmission device is used to mechanically connect with the controlled object, and is used to drive the controlled object to run according to the force of the follow-up system, the controlled object being the barrel.
