System and method for researching dynamic response of free target under action of underwater near-field explosion

Abstract

The invention provides a free target dynamic response research system and method under the action of underwater near-field explosion, belongs to the technical field of underwater explosion experiments, and aims to improve the dynamic response acquisition precision of a free target under the action of underwater near-field explosion. The system controls release of a free target through an electromagnet, combines interaction of bubbles generated by electric spark discharge and the free target, and adopts a high-speed camera and a data acquisition device to synchronously capture a response process. The system comprises an electromagnet, a slotted iron rod, a positioning trough plate, a glass water tank, a light source, a high-speed camera and the like, and can realize accurate and synchronous data acquisition and non-contact parameter extraction. A YOLOv8 model is used for cylinder detection, displacement, speed and acceleration are calculated, and a bubble area in an image is segmented through a U-Net model. The system can provide a reliable platform for underwater explosion dynamics research, and further development of the field is effectively promoted.

Description

Technical Field

[0001] The present invention belongs to the technical field of underwater explosion experiments, and in particular relates to a system and method for studying the dynamic response of a free target under the action of an underwater near-field explosion. Background Art

[0002] Currently, most existing underwater explosion test equipment focuses on measuring the pressure of fixed structures under near-field underwater explosions, while the dynamic response of free targets remains significantly limited. Existing technologies primarily rely on conducting experiments on fixed structures, measuring pressure changes under the impact of the explosion, and simulating and extrapolating these changes through methods such as finite element analysis. However, the dynamic behavior of free targets under near-field underwater explosions is complex, particularly the transient impact of the bubble secondary jet on the target. These effects involve multi-dimensional responses of the free target, including displacement and velocity, and differ significantly from the response mechanisms of fixed structures.

[0003] Existing experimental setups cannot adequately simulate the motion and response of free targets in explosive environments, and lack effective free target placement methods and precise dynamic response monitoring methods. Furthermore, while the dynamic response of free targets to underwater near-field explosions has important applications in fields such as marine engineering, military protection, and deep-sea exploration, existing technologies have yet to develop a comprehensive, systematic collection and analysis method for this purpose.

[0004] Existing research largely relies on single data acquisition methods, such as pressure sensors or accelerometers, lacking a comprehensive, multi-angle, real-time monitoring and analysis system for the target's overall dynamic response. This hinders in-depth research on the dynamic response of free-standing targets under underwater explosions and hinders reliable technical support for engineering practice. Therefore, the development of a system capable of efficiently acquiring the dynamic response of free-standing targets under underwater near-field explosions, particularly for analyzing the effects of bubble secondary jets, is urgently needed. Summary of the Invention

[0005] The purpose of the present invention is to provide a system and method for studying the dynamic response of a free target under the action of an underwater near-field explosion, so as to improve the acquisition accuracy of the dynamic response of a free target under the action of an underwater near-field explosion.

[0006] The purpose of the present invention is achieved through the following technical solutions:

[0007] A system for studying the dynamic response of a free target under an underwater near-field explosion comprises: an experimental stand, a three-coordinate controller mounted on the upper portion of the experimental stand, the three-coordinate controller being connected to a spark pedestal via a connecting rod, a discharge needle mounted on the spark pedestal, the spark pedestal and the discharge needle being both immersed below the water surface of a glass water tank, the glass water tank being mounted within the experimental stand;

[0008] A free target release mechanism is provided above the glass water tank, and the free target release mechanism is installed inside the experimental bracket;

[0009] The discharge box is connected to the discharge needle through a wire, and the discharge box is connected to a high-speed camera and a data acquisition device through a synchronization sensor. The discharge box controls the on and off of the electromagnet of the free target release mechanism by transmitting a signal to the electromagnet control box. The discharge box, the high-speed camera and the data acquisition device work synchronously, and the discharge of the discharge needle and the release of the free target cylinder of the free target release mechanism are carried out simultaneously.

[0010] Furthermore, the free target release mechanism includes an electromagnet, which is fixed on an electromagnet fixing plate, which is fixed on a positioning slot plate, which is fixed on the experimental bracket. When the electromagnet is energized, it attracts the slotted iron rod, which has slots on both sides and hangs a line on the slotted iron rod, and a free target cylinder with slots on both sides is hung below the line.

[0011] Furthermore, it also includes a light source, and the light source and the high-speed camera are placed on both sides of the experimental bracket.

[0012] Furthermore, the free target cylinder has a density of 1.3 g / cm 3 ABS plastic 3D printed cylinder with a filling rate of 77%.

[0013] The present invention may also include:

[0014] A method for studying the dynamic response of a free target under an underwater near-field explosion, using the above-mentioned system, comprises the following steps:

[0015] Step 1: Pour water into the glass water tank, then slowly lower the free target cylinder hanging under the line into the water, ensuring that the free target is suspended in the water near the center of the glass water tank under the action of the line;

[0016] Step 2: Adjust the longitudinal position of the spark pedestal according to the experimental content so that the free target cylinder and the discharge needle are located below the water surface and the distance between the free target cylinder and the discharge needle meets the experimental requirements; adjust the light source and high-speed camera so that the light intensity of the light source meets the experimental requirements and the shooting position of the light source and the high-speed camera are on the same horizontal line with the discharge needle;

[0017] Step 3: It is necessary to ensure that the electromagnet fails at the same time as the spark discharge, thereby releasing the free target hung on the slotted iron rod through the wire. There are components in the circuits of the discharge box and the electromagnet control box. The circuit board control time is set to 500ms to delay the spark discharge.

[0018] Step 4: Start the experiment, use the discharge box to make the underwater discharge needle generate underwater bubbles, the high-speed camera and data acquisition device work synchronously, and the electromagnet is disabled;

[0019] Step 5: Based on the 23,000fps high-definition video file captured by the high-speed camera, extract each frame from the video file and save it as an image. Customize the video path and output directory, read the video frame by frame, and save it as a .png file. Name each frame 1.png, 2.png, 3.png, ... N.png according to the order of the frames, ensuring that the image files are saved in order. The OpenCV library is used in the frame extraction process. The video file is opened with the cv2.VideoCapture() function, and the image is read frame by frame using cap.read(). Finally, the image is saved to the specified directory.

[0020] Step 6: Automatically detect cylinders from the image and calculate the ratio of pixels to millimeters based on the actual cylinder diameter. The calibration process uses the cv2.HoughCircles() function to achieve circle detection.

[0021] Step 7: Use YOLOv8 to train a cylinder detection model. Provide a dataset configuration file containing training images and their corresponding labels. By calling the YOLOv8 training script, the program loads the dataset and uses the pre-trained YOLOv8 weights for training. During the training process, the model is trained according to the configured parameters, including the number of training rounds, batch size, and image size. Finally, a weight file for cylinder detection is generated. The model can identify cylinders in the image and output their location and confidence level.

[0022] Step 8: Use the trained YOLOv8 model to detect cylinders in the new image sequence. By loading the trained weight file, reading the images frame by frame, and detecting them using the YOLOv8 model, the bounding box coordinates of the cylinder are obtained, the center position is calculated, and the pixel coordinates of the center are converted to actual millimeter coordinates using the calibration scale. In each frame of the image, the program will mark the center position of the cylinder and calculate the displacement, velocity, and acceleration. This information will be saved in a CSV file. At the same time, an image containing the annotation results will be generated and saved to the specified directory.

[0023] Step 9: Further process the prediction results, namely the displacement, velocity, and acceleration of the free target. Use the mutual information method and autocorrelation method to calculate the time delay. Transform the one-dimensional time series into a high-dimensional phase space through delay embedding. Plot the phase space trajectory to show the dynamic behavior of the time series in the phase space. View the results through a visual image file.

[0024] Step 10: Select part of the original image saved in step 5, circle the bubbles with red circles, then read the image, convert the image to the HSV color gamut, extract the red area, and generate a binary mask of the red area by setting the upper and lower limits of red. Combine the mask with the grayscale image to extract the bubble area. Finally, use the pad_or_crop_to_512 function to resize the mask to 512x512 to ensure a uniform input size for subsequent model training.

[0025] Step 11: Use the U-Net model for image segmentation training. Load the image and corresponding mask from the specified directory, resize the image to 512×512, and normalize it. Train the U-Net model using the training data and optimize the model parameters to accurately identify the bubble area. The training process uses the cross entropy loss function and updates the parameters using the Adam optimizer. After training, the model weights will be saved for subsequent predictions.

[0026] Step 12: Use the trained U-Net model to predict the new image and mark the bubble locations in the image. After resizing all the images saved in step 5 to 512×512, use the model to predict and generate a mask of the bubble area. Then, extract the bubble outline and mark the bubble location on the original image. Finally, save the annotated image and prediction results to the specified directory.

[0027] Furthermore, in step 6, the cylinder is automatically detected from the image, and the ratio of pixels to millimeters is calculated based on the actual cylinder diameter. The calibration process implements circle detection through the cv2.HoughCircles() function, including the following steps:

[0028] First, convert the image to grayscale and perform Gaussian blur processing;

[0029] Then, circles are detected by Hough transform, and the circle with the smallest distance to the image center is selected as the calibration object;

[0030] Finally, the program calculates the ratio between the circle's pixel diameter and its actual diameter and saves the ratio to a file, providing a scale for subsequent cylinder detection and displacement calculation.

[0031] Furthermore, in step 9, the dynamic behavior in the phase space of the dynamic response of the free target under the action of the underwater near-field explosion is calculated, and the calculation process includes: selecting a time delay τ, and calculating the optimal time delay by the mutual information method or the autocorrelation method; selecting an embedding dimension m, and using the FalseNearestNeighbors method to determine the appropriate dimension; then performing phase space reconstruction, and using time delay embedding to convert the one-dimensional time series into a trajectory in the m-dimensional phase space; then calculating the Lyapunov exponent, and analyzing the chaos of the system by tracking the perturbation trajectory; finally, calculating the correlation dimension, and solving the complexity of the phase space by calculating the distance between point pairs in the phase space.

[0032] Furthermore, the time delay τ is selected, and the optimal time delay τ is found by calculating the mutual information to maximize the mutual information MI(X t ,X t+τ ) is the optimal time delay, where:

[0033] MI(X t ,X t+τ )=H(X t )+H(X t+τ )-H(X t ,X t+τ ) (1)

[0034] Where: H(X t ) and H(X t+τ ) is the entropy of the time series at time t and t+τ, which measures the distribution of the data; H(X t ,X t+τ ) is X t and X t+τ The joint entropy of , which measures their dependencies;

[0035] The lag τ that maximizes the mutual information is selected as the optimal time delay:

[0036]

[0037] The autocorrelation function measures the similarity between a time series and its lagged version. The autocorrelation function is calculated as follows:

[0038]

[0039] Where: μ is the mean of the time series; τ is the time delay; X t and X t+τ are the values ​​of the time series at time t and t+τ respectively;

[0040] The optimal time delay τ is chosen by finding the position where the autocorrelation first falls below a threshold:

[0041]

[0042] The process of phase space reconstruction transforms the one-dimensional time series into a trajectory in a high-dimensional phase space through time-delay embedding;

[0043] For a given time delay τ and embedding dimension m, each point in the phase space can be represented as a vector:

[0044] x t =(X t ,X t+τ ,X t+2τ ,…,X t+(m-1)τ ) (5)

[0045] Where: x t is the embedding vector at time t, which contains multiple lagged values ​​of the time series; τ is the time delay, which determines the time interval between adjacent points; m is the embedding dimension, which determines the dimension of the phase space; by arranging these embedding vectors in chronological order, we can get the phase space trajectory.

[0046] Furthermore, the Lyapunov exponent quantifies the sensitivity of the phase space trajectory to the initial conditions, that is, the expansion speed of the trajectory, which is measured by calculating the exponential growth rate of the trajectory perturbation. The Lyapunov exponent calculation formula is:

[0047]

[0048] Where δx(t) is the distance between the perturbation trajectory and the original trajectory starting from the initial condition x(0); λ is the Lyapunov exponent, which indicates the stability of the system; if λ>0, the system is chaotic, and if λ<0, the system is stable.

[0049] Furthermore, the correlation dimension is used to quantify the complexity of the phase space and describe the dimensionality of the system state space. The calculation formula is as follows:

[0050]

[0051] Where: D2 is the correlation dimension, which represents the roughness or complexity of the phase space; It is the number of all point pairs in the phase space whose distance is less than a certain threshold ∈; ∈ is the distance threshold, which is calculated by gradually reducing the threshold.

[0052] The beneficial effects of the present invention are:

[0053] The experimental device of the present invention has significant advantages in improving the scientificity and applicability of the research on the dynamic response of free targets under the action of underwater near-field explosions, and lays a solid foundation for in-depth research on the impact mechanism of underwater explosions on free targets. Through modular design and high flexibility, the system can achieve precise and synchronous data acquisition, non-contact parameter extraction and intelligent nonlinear dynamic behavior analysis, and adapt to complex experimental environments and various experimental needs; the device not only provides a reliable experimental platform for the study of the dynamic response of free targets, but also provides technical support for related applications in underwater engineering, military protection and deep-sea exploration, and promotes the further development of the field of underwater explosions.

[0054] The system uses electromagnets to control the release of a free target, combining the interaction between bubbles generated by spark discharge and the free target with a high-speed camera and data acquisition device to synchronously capture the response process. This system enables precise and synchronized data acquisition and contactless parameter extraction. The YOLOv8 model is used for cylinder detection, calculating displacement, velocity, and acceleration, and a U-Net model is used to segment bubble regions within the image. This system provides a reliable platform for underwater explosion dynamics research and effectively promotes further development in this field. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Attachment Figure 1 It is an overall schematic diagram of the system for studying the dynamic response of a free target under the action of an underwater near-field explosion in the present invention.

[0056] Attachment Figure 2 It is a schematic diagram of the acquisition system and the electronic control system in the present invention.

[0057] Attachment Figure 3 It is a schematic diagram of the installation method of the free target release mechanism of the present invention;

[0058] Attachment Figure 4 It is a schematic diagram of the free target release mechanism in the present invention.

[0059] Attachment Figure 5 It is a schematic diagram of the free target release mechanism and the electric spark discharge positioning mechanism in the present invention.

[0060] Attachment Figure 6 It is a cross-sectional view of the free target in the present invention.

[0061] Attachment Figure 7 It is a schematic diagram of the connection between the discharge box and the electromagnet control box components in the present invention.

[0062] Attachment Figure 8 It is a schematic diagram of the bubble mask in the present invention.

[0063] Attachment Figure 9 It is a schematic diagram of the bubble area marking in the present invention.

[0064] Attachment Figure 10 It is a schematic diagram of the center mark of the free target side in the present invention.

[0065] Attachment Figure 11 It is a schematic diagram of the displacement, velocity and acceleration of a free target.

[0066] In the attached figure: 1. electromagnet, 2. electromagnet fixing plate, 3. slotted iron rod, 4. wire, 5. positioning slot plate, 6. free target cylinder, 7. discharge needle, 8. spark base, 9. connecting rod, 10. three-coordinate controller, 10-1. Z-axis motor, 10-2. X-axis motor, 10-3. Y-axis motor, 11. fastening bolt, 12. experimental bracket, 13. data acquisition device, 14. electromagnet control box, 15. discharge box, 16. glass water tank, 17. light source, 18. high-speed camera. DETAILED DESCRIPTION

[0067] The present invention will be further described below with reference to the accompanying drawings.

[0068] Example 1:

[0069] The present invention provides a free target dynamic response research system under underwater near-field explosion. Figure 1-6 As shown, it includes: an experimental bracket 12, a three-coordinate controller 10 is installed on the upper part of the experimental bracket 12, the three-coordinate controller 10 is connected to the spark base 8 through a connecting rod 9, a discharge needle 7 is installed on the spark base, the spark base 8 and the discharge needle 7 are all immersed below the water surface in the glass water tank 16, and the glass water tank 16 is installed in the experimental bracket;

[0070] A free target release mechanism is provided above the glass water tank 16 , and the free target release mechanism is installed inside the experimental bracket 12 ;

[0071] The discharge box 15 is connected to the discharge needle 7 through a wire, and the discharge box 15 is connected to the high-speed camera 18 and the data acquisition device 13 through a synchronization sensor. The discharge box 15 controls the on and off of the electromagnet 1 of the free target release mechanism by transmitting a signal to the electromagnet control box 14. The discharge box 15, the high-speed camera 18 and the data acquisition device 13 work synchronously, and the discharge of the discharge needle 7 and the release of the free target cylinder 6 of the free target release mechanism are carried out simultaneously.

[0072] The discharge box 15 is linked to the electromagnet control box 14 circuit to ensure that the electromagnet 1 releases the free target cylinder 6 and the discharge of the discharge needle 7 is synchronized, and the high-speed camera 18 and the data acquisition device 13 cooperate to capture the dynamic response process of the free target cylinder 6.

[0073] This embodiment further includes a light source 17 , and the light source 17 and the high-speed camera 18 are placed on both sides of the experimental support 12 .

[0074] The system experimental method of the present invention includes: filling water into a glass water tank 16 and suspending a free target cylinder 6; adjusting the position of the electric spark base 8 and the free target cylinder 6, and optimizing the shooting angles of the light source 17 and the high-speed camera 18; releasing underwater bubbles through the discharge needle 7 and simultaneously releasing the free target cylinder 6 to capture the interaction process; and finally analyzing the experimental data frame by frame and calculating the dynamic parameters.

[0075] The YOLOv8 model is used to detect images taken by the high-speed camera 18, calculate the position, displacement, velocity and acceleration of the free target cylinder 6, and store the data as a CSV file; the U-Net model is used to perform bubble segmentation, accurately mark the bubble area, and optimize the experimental data analysis.

[0076] The YOLOv8 model training accuracy reaches 95%, the displacement accuracy is 0.1 mm / frame, and the velocity accuracy is 0.2 mm / s, ensuring high-precision free-range target cylindrical 6 motion trajectory measurement.

[0077] OpenCV is used to implement image frame extraction and Hough circle detection, automatically calibrate the free target cylinder 6 and perform displacement calculation.

[0078] The system adopts a modular design to ensure precise synchronization of data acquisition and free target release, and realize the study of the dynamic response of free targets in underwater explosion environments.

[0079] As attached Figure 4 As shown, the free target release mechanism includes an electromagnet 1, which is fixed on an electromagnet fixing plate 2, which is fixed on a positioning slot plate 5, which is fixed on the experimental bracket 12. After the electromagnet 1 is energized, it attracts the slotted iron rod 3, which has slots on both sides and hangs a wire 4 on the slotted iron rod 3, and a free target cylinder 6 with slots on both sides is hung below the wire 4.

[0080] As attached Figure 5 As shown, the three-coordinate controller 10 includes a Z-axis motor 10 - 1 , an X-axis motor 10 - 2 , and a Y-axis motor 10 - 3 .

[0081] As attached Figure 6 As shown, the free target cylinder 6 has a density of 1.3 g / cm 3 The cylinder is 3D printed from ABS plastic with a filling rate of 77%, ensuring that the gravity is slightly greater than the buoyancy.

[0082] In this embodiment, the free target cylinder 6 will produce a dynamic response when subjected to a near-field explosion underwater. If the attack intensity is large, the free target cylinder 6 may even be damaged. The attack process is simulated in a glass water tank 16, and the free target cylinder 6 is set to be a cylinder with grooves.

[0083] Slowly lower the free target cylinder 6 hanging below the line into the water, ensuring that the free target is suspended in the water close to the center of the glass water tank under the action of the line.

[0084] When an underwater near-field spark explosion load acts on the surface of free target cylinder 6, free target cylinder 6 will exhibit a corresponding response, including movement and damage. By adjusting the mounting position of spark base 8, the discharge needle 7 and free target cylinder 6 can maintain a certain relative position. Discharge needle 7 is mounted on spark base 8, which is mounted on a three-dimensional coordinate controller 10. Both discharge needle 7 and spark base 8 are submerged below the water surface in glass water tank 16. The position of free target cylinder 6 within glass water tank 16 can be adjusted by varying the length of wire 4.

[0085] In this embodiment, the discharge box 15 is connected to the discharge needle 7 through a wire, and the discharge box 15 is connected to the high-speed camera 18 and the data acquisition device 13 through a synchronization sensor. The purpose is to ensure that the discharge box 15, the high-speed camera 18, and the data acquisition device 13 work synchronously. Because the time from the generation to the collapse of the underwater explosion bubble is very short (millisecond level), it is necessary for each device to work simultaneously to avoid missing relevant data and phenomena due to the short time of the experimental phenomenon. In particular, it is necessary to ensure that the discharge box 15 and the electromagnet control box 14 work simultaneously, and the generation of the spark bubble and the free target cylinder 6 obtaining a free state occur simultaneously, so as to capture the process from the spark bubble to the free target cylinder 6 starting to contact until collapse.

[0086] The experimental use environment of the present invention is as shown in the attached Figure 1 As shown, the glass water tank 16 is made of transparent glass to ensure that the light emitted by the light source 17 can pass through the glass water tank 16. At the same time, the high-speed camera 18 collects the movement status of the bubbles and the free target 6 in the glass water tank 16. During the experiment, the glass water tank 16 is placed on the experimental stand and fixed. The light source 17 and the high-speed camera 18 are placed on both sides of the glass water tank 16 respectively. The light source 17 is 0.5 meters away from the wall of the glass water tank 16. The high-speed camera 18 is 0.5 meters away from the wall of the glass water tank 16. The center of the light source 17 is at the same height as the high-speed camera. The experimental table is placed approximately 2 meters away from the glass water tank 16. The data acquisition device 13 and the discharge box 15 are placed on the experimental table.

[0087] In this embodiment, the experimental bracket 12 is composed of aluminum profiles as a whole, and the aluminum profiles are connected by fastening bolts 11 and angle brackets. The experimental bracket is 1.5m high and 0.78m wide. The overall frame of the bracket is composed of 4 1.5m and 8 0.7m aluminum profiles. The glass water tank 16 platform is spliced ​​with 7 0.7m long aluminum profiles and is installed in the lower part of the overall structure of the experimental bracket. The upper platform is 0.5m above the ground.

[0088] During the experiment, the glass water tank 16 is installed and fixed on the glass water tank 16 platform of the experimental bracket; the glass water tank 16 is adjusted to the middle position; the beam is fixed on the experimental bracket; the structural frame and the beam are fixed by tightening the bolts 11; the height of the structural frame is adjusted so that the free target 6 is in the center position below the horizontal plane inside the glass water tank 16.

[0089] Adjust the distance between the discharge needle 7 and the free target 6; the discharge needle 7 generates electric sparks, which in turn causes the discharge needle 7 in the glass water tank 16 to generate bubbles. The bubbles couple with the free target 6 and interact with each other. Under the interaction, the free target 6 produces a dynamic response; at the same time, the bubbles pulsate and collapse; the high-speed camera 18 records the entire experimental process, and the data acquisition device 13 stores the video file of the experimental process.

[0090] The high-speed camera 18 is used in conjunction with the data acquisition device 13 to capture the contact process between the electric spark bubble and the free target 6, simulating the attack process of the underwater near-field explosion, and providing an experimental basis for analyzing the near-field bubble pulsation and load characteristics at the boundary of the free target 6 under different attack angles, different explosion distances and other working conditions.

[0091] As attached Figure 7 As shown, the discharge box 15 and the electromagnet control box 14 include two main power switches, the two main power switches receive 220V AC power respectively, and the two main power switches output 24V DC power respectively, one of which is connected to the electromagnet intermediate relay, and the electromagnet intermediate relay is connected to the electromagnet interface. The other 24V DC power supply is connected to the charge-discharge switch, the circuit board, and the fan respectively, the charge-discharge switch is connected to the authorization switch, the authorization switch is connected to the discharge switch, the discharge switch is connected to the circuit board, the circuit board is connected to the discharge relay, the discharge relay is connected to 220V AC power and the vacuum contactor, the charge-discharge switch is connected to the charging relay, the charging relay is connected to the charging input interface and the energy storage capacitor module respectively, the energy storage capacitor module is connected to the vacuum contactor and the voltmeter, and the vacuum contactor is connected to the discharge output interface.

[0092] Example 2:

[0093] According to the system for studying the dynamic response of a free target under an underwater near-field explosion described in Example 1, a method for studying the dynamic response of a free target under an underwater near-field explosion using the system includes the following steps:

[0094] Step 1: Pour water into the glass water tank 16, and then slowly lower the free target cylinder 6 hanging under the line 4 into the water, ensuring that the free target is suspended in the water near the center of the glass water tank 16 under the action of the line 4;

[0095] Step 2: Adjust the longitudinal position of the spark pedestal 8 according to the experimental content so that the free target cylinder 6 and the discharge needle 7 are located below the water surface and the distance between the free target cylinder 6 and the discharge needle 7 meets the experimental requirements; adjust the light source 17 and the high-speed camera 18 so that the illumination intensity of the light source 17 meets the experimental requirements and the shooting positions of the light source and the high-speed camera 18 are on the same horizontal line as the discharge needle 7;

[0096] Step 3: It is necessary to ensure that the electromagnet 7 fails at the same time as the spark discharge, thereby releasing the free target hung on the slotted iron rod 3 through the wire 4. There are components in the circuit of the discharge box 15 and the electromagnet control box 14, and the circuit board control time is set to delay the spark discharge by 500ms;

[0097] Step 4: Start the experiment, use the discharge box to make the underwater discharge needle generate underwater bubbles, the high-speed camera and the data acquisition device work synchronously, and the electromagnet 1 is disabled;

[0098] Step 5: Based on the 23,000fps high-definition video file captured by the high-speed camera, extract each frame from the video file and save it as an image. Customize the video path and output directory, read the video frame by frame, and save it as a .png file. Name each frame 1.png, 2.png, 3.png, ... N.png according to the order of the frames, ensuring that the image files are saved in order. The OpenCV library is used in the frame extraction process. The video file is opened with the cv2.VideoCapture() function, and the image is read frame by frame using cap.read(). Finally, the image is saved to the specified directory.

[0099] Step 6: Automatically detect cylinders from the image and calculate the ratio of pixels to millimeters based on the actual cylinder diameter. The calibration process uses the cv2.HoughCircles() function to achieve circle detection.

[0100] Step 7: Use YOLOv8 to train a cylinder detection model. Provide a dataset configuration file containing training images and their corresponding labels. By calling the YOLOv8 training script, the program loads the dataset and uses the pre-trained YOLOv8 weights for training. During the training process, the model is trained according to the configured parameters, including the number of training rounds, batch size, and image size. Finally, a weight file for cylinder detection is generated. The model can identify cylinders in the image and output their location and confidence level.

[0101] Step 8: Use the trained YOLOv8 model to detect cylinders in the new image sequence. By loading the trained weight file, reading the image frame by frame, and detecting it using the YOLOv8 model, we can get the bounding box coordinates of the cylinder, calculate the center position, and convert the pixel coordinates of the center of the circle into actual millimeter coordinates through the calibration ratio. In each frame of the image, the program will mark the center position of the cylinder. The marked cylinder position is shown in the attached figure. Figure 10 As shown, the displacement, velocity and acceleration are calculated and saved in a CSV file. An image containing the annotation results is generated and saved to the specified directory.

[0102] Step 9: Further process the prediction results, which are the displacement, velocity and acceleration of the free target. Use the mutual information method and autocorrelation method to calculate the time delay. Transform the one-dimensional time series into a high-dimensional phase space through delay embedding, draw the phase space trajectory, and display the dynamic behavior of the time series in the phase space. View the results through a visual image file. An example of the visual result is shown in the attached figure. Figure 11 As shown;

[0103] Step 10: Select part of the original image saved in step 5, circle the bubbles with red circles, then read the image, convert the image into HSV color space, extract the red area, and generate a binary mask of the red area by setting the upper and lower limits of red. The mask form is as shown in the attached figure. Figure 8 As shown in the figure, the mask is combined with the grayscale image to extract the bubble area. Finally, the mask is resized to 512×512 using the pad_or_crop_to_512 function to ensure a uniform input size for subsequent model training.

[0104] Step 11: Use the U-Net model for image segmentation training. Load the image and corresponding mask from the specified directory, resize the image to 512×512, and normalize it. Train the U-Net model using the training data and optimize the model parameters to accurately identify the bubble area. The training process uses the cross entropy loss function and updates the parameters using the Adam optimizer. After training, the model weights will be saved for subsequent predictions.

[0105] Step 12: Use the trained U-Net model to predict the new image and mark the bubble positions in the image; resize all the images saved in step 5 to 512×512, use the model to predict and generate a mask of the bubble area, then extract the bubble outline and mark the bubble position on the original image. The marked bubble position is shown in the figure below. Figure 9 As shown, finally the labeled images and prediction results are saved to the specified directory.

[0106] Furthermore, in step 6, the cylinder is automatically detected from the image, and the ratio of pixels to millimeters is calculated based on the actual cylinder diameter. The calibration process implements circle detection through the cv2.HoughCircles() function, including the following steps:

[0107] First, convert the image to grayscale and perform Gaussian blur processing;

[0108] Then, circles are detected by Hough transform, and the circle with the smallest distance to the image center is selected as the calibration object;

[0109] Finally, the program calculates the ratio between the circle's pixel diameter and its actual diameter and saves the ratio to a file, providing a scale for subsequent cylinder detection and displacement calculation.

[0110] Furthermore, in step 9, the dynamic behavior in the phase space of the dynamic response of the free target under the action of the underwater near-field explosion is calculated, and the calculation process includes: selecting a time delay τ and calculating the optimal time delay by using the mutual information method or the autocorrelation method; selecting an embedding dimension m and using the False Nearest Neighbors method to determine the appropriate dimension; then performing phase space reconstruction and using time delay embedding to convert the one-dimensional time series into a trajectory in the m-dimensional phase space; then calculating the Lyapunov exponent and analyzing the chaos of the system by tracking the perturbation trajectory; and finally calculating the correlation dimension and solving the complexity of the phase space by calculating the distance between point pairs in the phase space.

[0111] Furthermore, the time delay τ is selected, and the optimal time delay τ is found by calculating the mutual information to maximize the mutual information MI(X t ,X t+τ ) is the optimal time delay, where:

[0112] MI(X t ,X t+τ )=H(X t )+H(X t+τ )-H(X t ,X t+τ ) (1)

[0113] Where: H(X t ) and H(X t+τ ) is the entropy of the time series at time t and t+τ, which measures the distribution of the data; H(X t ,X t+τ ) is X t and X t+τ The joint entropy of , which measures their dependencies;

[0114] The lag τ that maximizes the mutual information is selected as the optimal time delay:

[0115]

[0116] The autocorrelation function measures the similarity between a time series and its lagged version. The autocorrelation function is calculated as follows:

[0117]

[0118] Where: μ is the mean of the time series; τ is the time delay; X t and X t+τ are the values ​​of the time series at time t and t+τ respectively;

[0119] The optimal time delay τ is chosen by finding the position where the autocorrelation first falls below a threshold:

[0120]

[0121] The process of phase space reconstruction transforms the one-dimensional time series into a trajectory in a high-dimensional phase space through time-delay embedding;

[0122] For a given time delay τ and embedding dimension m, each point in the phase space can be represented as a vector:

[0123] x t =(X t ,X t+τ ,X t+2τ ,…,X t+(m-1)τ ) (5)

[0124] Where: x t is the embedding vector at time t, which contains multiple lagged values ​​of the time series; τ is the time delay, which determines the time interval between adjacent points; m is the embedding dimension, which determines the dimension of the phase space; by arranging these embedding vectors in chronological order, we can get the phase space trajectory.

[0125] Furthermore, the Lyapunov exponent quantifies the sensitivity of the phase space trajectory to the initial conditions, that is, the expansion speed of the trajectory, which is measured by calculating the exponential growth rate of the trajectory perturbation. The Lyapunov exponent calculation formula is:

[0126]

[0127] Where δx(t) is the distance between the perturbation trajectory and the original trajectory starting from the initial condition x(0); λ is the Lyapunov exponent, which indicates the stability of the system; if λ>0, the system is chaotic, and if λ<0, the system is stable.

[0128] Furthermore, the correlation dimension is used to quantify the complexity of the phase space and describe the dimensionality of the system state space. The calculation formula is as follows:

[0129]

[0130] Where: D2 is the correlation dimension, which represents the roughness or complexity of the phase space; It is the number of all point pairs in the phase space whose distance is less than a certain threshold ∈; ∈ is the distance threshold, which is calculated by gradually reducing the threshold.

[0131] In this embodiment, the YOLOv8 model is used to detect cylinders based on the dynamic response captured by the high-speed camera 18, and the image is read frame by frame to calculate the position of the cylinder. By training the YOLOv8 model and generating a weight file, the model achieves an accuracy of 95% on the validation set, and is able to identify cylinders in the image and output the position and confidence level, with a confidence level of 0.85. Based on the ratio between the actual diameter of 10 mm and the pixel, the ratio of pixels to millimeters is 0.5, and the pixel coordinates of the cylinder are converted to actual millimeter coordinates. The displacement, velocity, and acceleration of the target are further calculated, with a displacement accuracy of 0.1 mm / frame and a velocity accuracy of 0.2 mm / second. All calculation results will be saved in a CSV file for subsequent data analysis.

[0132] In this example, the bubble region in the image is segmented using a U-Net model. Image preprocessing is first performed, and the image is resized to 512x512 and normalized using a normalization range of 0-1. The U-Net model is then trained using a cross-entropy loss function. The Adam optimizer is used during training, with a learning rate set to 0.001. Training is performed for 50 epochs, resulting in a Dice coefficient of 0.92 and an accuracy of 90% on the validation set. After training, the U-Net model can predict new images, mark the bubble locations, and annotate them on the original image. Finally, an annotated image containing the bubble region and the prediction results are generated and saved to a specified directory for subsequent analysis.

[0133] The experimental device of the present invention, through its modular design and high flexibility, can accurately and synchronously collect data, extract parameters non-contactly, and perform intelligent dynamic analysis, significantly improving the scientificity and applicability of the study of the dynamic response of free targets under underwater near-field explosions, and providing a reliable experimental platform.

[0134] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A system for studying the dynamic response of a free target under underwater near-field explosion, characterized in that: include: An experimental support (12), wherein a three-coordinate controller (10) is installed on the upper portion of the experimental support (12), wherein the three-coordinate controller (10) is connected to an electric spark base (8) via a connecting rod (9), wherein a discharge needle (7) is installed on the electric spark base, wherein the electric spark base (8) and the discharge needle (7) are all immersed below the water surface in a glass water tank (16), and wherein the glass water tank (16) is installed in the experimental support; A free target release mechanism is provided above the glass water tank (16), and the free target release mechanism is installed inside the experimental support (12); The discharge box (15) is connected to the discharge needle (7) via a wire. The discharge box (15) is connected to a high-speed camera (18) and a data acquisition device (13) via a synchronization sensor. The discharge box (15) controls the on and off of the electromagnet (1) of the free target release mechanism by transmitting a signal to an electromagnet control box (14). The discharge box (15), the high-speed camera (18) and the data acquisition device (13) work synchronously. The discharge of the discharge needle (7) and the release of the free target cylinder (6) of the free target release mechanism are carried out simultaneously.

2. The free target dynamic response research system under underwater near-field explosion according to claim 1 is characterized in that: The free target release mechanism comprises an electromagnet (1), wherein the electromagnet (1) is fixed on an electromagnet fixing plate (2), the electromagnet fixing plate (2) is fixed on a positioning slot plate (5), and the positioning slot plate (5) is fixed on the experimental bracket (12). When the electromagnet (1) is energized, it attracts a slotted iron rod (3), the slotted iron rod (3) has slots on both sides, and a wire (4) is hung on the slotted iron rod (3), and a free target cylinder (6) with slots on both sides is hung below the wire (4).

3. The free target dynamic response research system under underwater near-field explosion according to claim 1 is characterized in that: The device further comprises a light source (17), wherein the light source (17) and the high-speed camera (18) are placed on both sides of the experimental support (12).

4. The free target dynamic response research system under underwater near-field explosion according to claim 1 is characterized in that: The free target cylinder (6) has a density of 1.3 g / cm 3 ABS plastic 3D printed cylinder with a filling rate of 77%.

5. A method for studying the dynamic response of a free target under underwater near-field explosion, the method using the system according to any one of claims 1 to 4, characterized in that: The following steps are involved: Step 1: Pour water into the glass water tank (16), and then slowly place the free target cylinder (6) hanging below the line (4) into the water, ensuring that the free target is suspended in the water near the center of the glass water tank (16) under the action of the line (4); Step 2: According to the experimental content, the longitudinal position of the spark base (8) is adjusted so that the free target cylinder (6) and the discharge needle (7) are located below the water surface, and the distance between the free target cylinder (6) and the discharge needle (7) meets the experimental requirements; the light source (17) and the high-speed camera (18) are adjusted so that the illumination intensity of the light source (17) meets the experimental requirements, and the shooting positions of the light source and the high-speed camera (18) are on the same horizontal line as the discharge needle (7); Step 3: It is necessary to ensure that the electromagnet (7) fails at the same time as the spark discharge, thereby releasing the free target hung on the slotted iron rod (3) through the wire (4), and there are components in the circuit of the discharge box (15) and the electromagnet control box (14), and the circuit board control time delay of 500ms is set to the spark discharge; Step 4: Start the experiment, use the discharge box to make the underwater discharge needle generate underwater bubbles, the high-speed camera and the data acquisition device work synchronously, and the electromagnet (1) is disabled; Step 5: Based on the 23,000fps high-definition video file captured by the high-speed camera, extract each frame from the video file and save it as an image. Customize the video path and output directory, read the video frame by frame, and save it as a .png file. Name each frame 1.png, 2.png, 3.png, ... N.png according to the order of the frames, ensuring that the image files are saved in order. The OpenCV library is used in the frame extraction process. The video file is opened with the cv2.VideoCapture() function, and the image is read frame by frame using cap.read(). Finally, the image is saved to the specified directory. Step 6: Automatically detect cylinders from the image and calculate the ratio of pixels to millimeters based on the actual cylinder diameter. The calibration process uses the cv2.HoughCircles() function to achieve circle detection. Step 7: Use YOLOv8 to train a cylinder detection model. Provide a dataset configuration file containing training images and their corresponding labels. By calling the YOLOv8 training script, the program loads the dataset and uses the pre-trained YOLOv8 weights for training. During the training process, the model is trained according to the configured parameters, including the number of training rounds, batch size, and image size. Finally, a weight file for cylinder detection is generated. The model can identify cylinders in the image and output their location and confidence level. Step 8: Use the trained YOLOv8 model to detect cylinders in the new image sequence. By loading the trained weight file, reading the images frame by frame, and detecting them using the YOLOv8 model, the bounding box coordinates of the cylinder are obtained, the center position is calculated, and the pixel coordinates of the center are converted to actual millimeter coordinates using the calibration scale. In each frame of the image, the program will mark the center position of the cylinder and calculate the displacement, velocity, and acceleration. This information will be saved in a CSV file. At the same time, an image containing the annotation results will be generated and saved to the specified directory. Step 9: Further process the prediction results, namely the displacement, velocity, and acceleration of the free target. Use the mutual information method and autocorrelation method to calculate the time delay. Transform the one-dimensional time series into a high-dimensional phase space through delay embedding. Plot the phase space trajectory to show the dynamic behavior of the time series in the phase space. View the results through a visual image file. Step 10: Select part of the original image saved in step 5, circle the bubbles with red circles, then read the image, convert the image to the HSV color gamut, extract the red area, and generate a binary mask of the red area by setting the upper and lower limits of red. Combine the mask with the grayscale image to extract the bubble area. Finally, use the pad_or_crop_to_512 function to resize the mask to 512×512 to ensure a uniform input size for subsequent model training. Step 11: Use the U-Net model for image segmentation training. Load the image and corresponding mask from the specified directory, resize the image to 512×512, and normalize it. Train the U-Net model using the training data and optimize the model parameters to accurately identify the bubble area. The training process uses the cross entropy loss function and updates the parameters using the Adam optimizer. After training, the model weights will be saved for subsequent predictions. Step 12: Use the trained U-Net model to predict the new image and mark the bubble locations in the image. After resizing all the images saved in step 5 to 512×512, use the model to predict and generate a mask of the bubble area. Then, extract the bubble outline and mark the bubble location on the original image. Finally, save the annotated image and prediction results to the specified directory.

6. The system for studying the dynamic response of a free target under underwater near-field explosion according to claim 5 is characterized in that: In step 6, cylinders are automatically detected from the image, and the ratio of pixels to millimeters is calculated based on the actual cylinder diameter. The calibration process uses the cv2.HoughCircles() function to implement circle detection, which includes the following steps: First, convert the image to grayscale and perform Gaussian blur processing; Then, circles are detected by Hough transform, and the circle with the smallest distance to the image center is selected as the calibration object; Finally, the program calculates the ratio between the circle's pixel diameter and its actual diameter and saves the ratio to a file, providing a scale for subsequent cylinder detection and displacement calculation.

7. The free target dynamic response research system under underwater near-field explosion according to claim 5 is characterized in that: In step 9, the dynamic behavior of the free target dynamic response in the phase space under the action of the underwater near-field explosion is calculated. The calculation process includes: selecting a time delay τ and calculating the optimal time delay by using the mutual information method or the autocorrelation method; selecting an embedding dimension m and determining the appropriate dimension by using the False Nearest Neighbors method; then performing phase space reconstruction and converting the one-dimensional time series into a trajectory in the m-dimensional phase space by using time delay embedding; then calculating the Lyapunov exponent and analyzing the chaos of the system by tracking the disturbance trajectory; and finally calculating the correlation dimension and solving the complexity of the phase space by calculating the distance between point pairs in the phase space.

8. The free target dynamic response research system under underwater near-field explosion according to claim 5 is characterized in that: The time delay τ is selected, and the optimal time delay τ is found by calculating the mutual information, maximizing the mutual information MI(X t ,X t+τ ) is the optimal time delay, where: MI(X t ,X t+τ )=H(X t )+H(X t+τ )-H(X t ,X t+τ ) (1) Where: H(X t ) and H(X t+τ ) is the entropy of the time series at time t and t+τ, which measures the distribution of the data; H(X t ,X t+τ ) is X t and X t+τ The joint entropy of , which measures their dependencies; The lag τ that maximizes the mutual information is selected as the optimal time delay: The autocorrelation function measures the similarity between a time series and its lagged version. The autocorrelation function is calculated as follows: Where: μ is the mean of the time series; τ is the time delay; X t and X t+τ are the values ​​of the time series at time t and t+τ respectively; The optimal time delay τ is chosen by finding the position where the autocorrelation first falls below a threshold: The process of phase space reconstruction transforms the one-dimensional time series into a trajectory in a high-dimensional phase space through time-delay embedding; For a given time delay τ and embedding dimension m, each point in the phase space can be represented as a vector: x t =(X t ,X t+τ ,X t+2τ ,…,X t+(m-1)τ ) (5) Where: x t is the embedding vector at time t, which contains multiple lagged values ​​of the time series; τ is the time delay, which determines the time interval between adjacent points; m is the embedding dimension, which determines the dimension of the phase space; by arranging these embedding vectors in chronological order, we can get the phase space trajectory.

9. The free target dynamic response research system under underwater near-field explosion according to claim 5 is characterized in that: The Lyapunov exponent quantifies the sensitivity of the phase space trajectory to the initial conditions, that is, the expansion speed of the trajectory, which is measured by calculating the exponential growth rate of the trajectory perturbation. The Lyapunov exponent calculation formula is: Where δx(t) is the distance between the perturbation trajectory and the original trajectory starting from the initial condition x(0); λ is the Lyapunov exponent, which indicates the stability of the system; if λ>0, the system is chaotic, and if λ<0, the system is stable.

10. The free target dynamic response research system under underwater near-field explosion according to claim 5, characterized in that: The correlation dimension is used to quantify the complexity of the phase space and describe the dimension of the system state space. The calculation formula is as follows: Where: D2 is the correlation dimension, which represents the roughness or complexity of the phase space; C(ò) is the number of all point pairs in the phase space whose distance is less than a certain threshold ∈; ∈ is the distance threshold, which is calculated by gradually reducing the threshold.

Images