A system and method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water

Abstract

The present invention discloses a system and method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water, which belongs to the technical field of submarine-launched UAVs. The evaluation system includes a submarine-launched UAV sensor system, a submarine-launched UAV communication system, a data processing and storage system, a CFD simulation analysis system, and a submarine-launched UAV entry and exit characteristics evaluation platform, which are arranged on a mother ship. The evaluation method uses a variety of sensors, CFD simulation, convolutional neural networks, bidirectional gated cyclic units, and SAO snow ablation hyperparameter optimization methods onboard the submarine-launched UAV to determine whether the entry and exit tasks can be completed as expected.

Description

Technical Field

[0001] The present invention belongs to the technical field of submarine-launched UAVs, and in particular relates to a system and method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water. Background Art

[0002] Submarines have advantages such as high concealment and long endurance, while drones have advantages such as high flexibility, low cost and remote control. Combining the two can enhance the submarine's reconnaissance capability, expand the submarine's operating range and increase the diversity of its tasks. We call drones that can be launched from submarines and perform missions submarine-launched drones.

[0003] The form of the submarine-launched UAV's mission is to be launched from the submarine's launch device, cross the water-air interface, and fly in the air to perform the mission. Since the submarine-launched UAV is accompanied by complex flow field motion and stress changes during the cross-medium process, it is necessary to ensure the structural and functional integrity of the submarine-launched UAV after entering and exiting the water and improve the mission success rate. The hydrodynamic characteristics analysis of the submarine-launched UAV's entry and exit process is crucial.

[0004] At present, the entry and exit of submarine-launched UAVs is in a stage of continuous updating and iteration. Generally speaking, the technical methods for studying the hydrodynamic characteristics of the entry and exit process of submarine-launched UAVs mainly include CFD numerical simulation, experimental testing and other means. The accuracy of the data obtained by numerical simulation alone is difficult to guarantee, and usually requires verification by means of experiments and other means before it can be used. If detailed entry and exit process data is obtained through experimental testing, the experimental conditions will be limited, and it will be difficult to cover all actual operating conditions. At the same time, the workload of data collection and analysis is very large, requiring precise measurement and data processing technology. At the same time, combining CFD numerical simulation with experimental testing methods can improve the accuracy of numerical simulation, and data that is difficult to obtain in experiments can be obtained to reduce the burden of experimental testing. In this way, all data are authentic and reliable, and rich and detailed motion status information can be provided, which is more conducive to the subsequent provision of a comprehensive evaluation report.

[0005] By detecting and calculating the hydrodynamic characteristics of the submarine-launched UAV entering and exiting the water and designing a real-time evaluation method, we can obtain parameters such as the power required at the moment of entering and exiting the water and the posture of the UAV. The evaluation results can help us determine whether the submarine-launched UAV can successfully fly from underwater into the air, or vice versa. It can also foresee and solve potential problems, ensure the reliability of the UAV in various tasks, and design a more adaptable UAV so that it can successfully perform tasks under various environmental conditions. Summary of the invention

[0006] In view of the above problems existing in the prior art, the present invention proposes a system and method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit from water. The system and method have a reasonable design, solve the deficiencies of the prior art, and have good effects.

[0007] A hydrodynamic characteristics evaluation system for a submarine-launched UAV during its entry and exit into water, comprising a submarine-launched UAV sensing system, a submarine-launched UAV communication system, a data processing and storage system, a CFD simulation analysis system, and a submarine-launched UAV entry and exit characteristics evaluation platform, wherein the data processing and storage system, the CFD simulation analysis system, and the submarine-launched UAV entry and exit characteristics evaluation platform are arranged on a mother ship;

[0008] The submarine-launched UAV sensing system includes a flow field sensor, a Doppler velocimeter, an acceleration sensor, a temperature-salinity-depth sensor, a GPS, and a gyroscope, which collects data on the launch, exit, and lift-off of the submarine-launched UAV in real time;

[0009] The submarine-launched UAV communication system is used to upload data to the mother ship for processing and analysis;

[0010] The data processing and storage system is used to process data from the submarine-launched UAV sensor system and store the processed data in a database;

[0011] The CFD simulation analysis system is used to obtain data that is difficult for sensors to obtain and has a significant impact on the movement of submarine-launched UAVs;

[0012] The submarine-launched UAV water entry and exit characteristics evaluation platform is used to analyze the hydrodynamic characteristics of the submarine-launched UAV water entry and exit process, and determine whether the submarine-launched UAV can complete the water entry and exit tasks as expected, decide whether the submarine-launched UAV performs the mission, make corresponding mission adjustments, and remotely issue control instructions to the UAV.

[0013] Furthermore, the Doppler speed meter is used to monitor the movement speed of the submarine-launched UAV, the acceleration sensor is used to monitor the acceleration of the submarine-launched UAV during its movement; the GPS is used to obtain the position of the submarine-launched UAV; the gyroscope is used to monitor the posture data of the submarine-launched UAV; the flow field sensor is used to measure the flow field velocity and pressure; the temperature, salinity and depth sensor is used to provide water temperature, density, salinity and depth.

[0014] Furthermore, the data processing and storage system is used for signal noise reduction processing for subsequent analysis, weighted smoothing processing to obtain motion state data after removing accidental data mutations, extraction mutation processing to obtain motion state data of instantaneous changes in water entry and exit, and simple moving average processing to obtain the most recent stable motion state data of the data stream; feature extraction is performed on the data to obtain time domain feature parameters and frequency feature parameters, and then feature normalization is performed to extract mean, extreme value, and zero-crossing point data.

[0015] Furthermore, the CFD simulation and analysis system is used to create a submarine-launched UAV geometric model based on the data in the data storage system as the initial state parameters, and perform mesh division and mesh independence verification; set flow field boundary conditions, define the fluid domain and the position of the submarine-launched UAV; create a simulation physical model of the submarine-launched UAV, select appropriate fluid models, turbulence models, boundary conditions, and material properties; configure solver settings, including time step, convergence criteria, and number of iterations to meet simulation requirements; perform simulation iterative calculations, and view the calculation results through visualization tools, including hydrodynamic characteristic parameter data such as attached water mass, posture, and speed, as well as velocity cloud maps, pressure cloud maps, and water surface impact cloud maps during the entry and exit process.

[0016] A method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit of water, using a system for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit of water as described above, comprises the following steps:

[0017] Step 1: Obtain the speed, acceleration, position, posture data of the submarine-launched UAV relative to the water surface depth, flow field speed and density, pressure, water temperature, density, salinity and depth data through the submarine-launched UAV sensor system;

[0018] Step 2: The data collected in step 1 is transmitted to the mother ship through the submarine-launched UAV communication system;

[0019] Step 3: Process the data through the mother ship data processing and storage module, including singular value decomposition denoising, weighted smoothing, extraction mutation processing, and simple moving average processing to obtain data with different characteristics;

[0020] Step 4: Extract features from the processed data, use Fourier transform to obtain time domain feature parameters and frequency feature parameters, normalize the time domain feature parameters and frequency feature parameters, extract mean, extreme value, and zero-crossing point data, and store them in the database;

[0021] Step 5: Determine the initial angle of the drone relative to the water surface, initial velocity, relative water surface depth, flow field velocity and density required for the CFD simulation of the submarine-launched drone, retrieve relevant parameters from the database, and obtain the initial angle of the drone relative to the water surface from the posture data;

[0022] Step 6: Create a simulation physical model of the submarine-launched UAV, perform mesh division and mesh independence verification on the UAV, and select the best simulation network;

[0023] Step 7: Set the flow field boundary conditions and simulation physical model, perform iterative simulation calculations, and obtain the force variation data, velocity variation data, acceleration variation data, angle variation data, flow field evolution data, and corresponding execution results of the submarine-launched UAV entry and exit mission under different conditions. The execution results of the submarine-launched UAV entry and exit mission include successful entry and exit and failure to enter and exit the water;

[0024] Step 8: Divide the data in the database and the data obtained by CFD simulation into training set, validation set and test set;

[0025] Step 9: Construct a CNN-BiGRU-Attention evaluation model, including CNN units, BiGRU units, and Attention units;

[0026] Step 10: Use the training set and validation set to train the CNN-BiGRU-Attention evaluation model, and use the test set to test the trained model to determine whether the model's evaluation ability is optimal. If the model's evaluation ability is not optimal, adjust the relevant hyperparameters;

[0027] Step 11: After optimizing the hyperparameters, continue training the CNN-BiGRU-Attention evaluation model algorithm to obtain a trained CNN-BiGRU-Attention evaluation model;

[0028] Step 12: Evaluate the motion state parameters input by the submarine-launched UAV to determine whether the entry and exit mission can be completed as expected. If the entry and exit mission cannot be completed, control the motion state of the UAV to achieve the conditions for the UAV to enter and exit the water normally.

[0029] Furthermore, it is characterized in that the CNN unit consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer;

[0030] The mathematical expression of the convolutional layer is:

[0031] ;

[0032] in, For the The convolutional layer Input feature maps or feature data; For the The convolutional layers are located at Weight data of For the Convolutional layer The bias term corresponding to the feature map; As the activation function, the ReLU activation function is selected, and the mathematical expression is:

[0033] ;

[0034] The mathematical expression of the pooling layer is:

[0035] ;

[0036] In the formula, For the The first pooling layer Input feature maps or feature data, For the Pooling layer The weight of a feature map or feature data; is the pooling function; For the The pooling window of the pooling layer, For the Pooling layer The bias term corresponding to the feature map.

[0037] Furthermore, the BiGRU unit includes two GRU units with opposite propagation directions, and the calculation process of the GRU unit is:

[0038] ;

[0039] ;

[0040] ;

[0041] ;

[0042] In the formula, To reset the gate, To update the gate, is the candidate hidden layer state, is the hidden layer output, is the Sigmoid function, tanh is the activation function, , , are the weight matrices input to the reset gate, update gate, and candidate hidden state, respectively, and are used to map the input vector to the corresponding gate or state; , , are the weight matrices from the previous hidden state to the reset gate, update gate, and candidate hidden state, respectively, which are used to map the hidden state of the previous time step to the corresponding gate or state; is the input vector of the current time step t; The previous time step The hidden state vector of

[0043] The calculation process of BiGRU is:

[0044] ;

[0045] ;

[0046] ;

[0047] In the formula, GRU is the calculation process of the above GRU unit, , are the forward and backward hidden layer outputs, , is the hidden layer output weight, for The hidden layer bias of .

[0048] Furthermore, the Attention unit is an attention mechanism that assigns different weight values ​​to feature data, focuses on key information, and highlights key features of the data. The mathematical expression is:

[0049] ;

[0050] In the formula, is the query vector, is the key vector, is a value vector.

[0051] Furthermore, in step 10, the SAO algorithm is introduced to optimize the hyperparameters, and the hyperparameters are regarded as molecular positions. The SAO algorithm process is as follows:

[0052] Step 10.1: Initialize the algorithm and generate a random population of molecules. The molecular position expression is:

[0053] ;

[0054] In the formula, For the Molecules Dimensional location; , are the upper and lower limits of the molecular position; is a random number in the interval [0,1]; is the number of molecules; is the number of hyperparameters;

[0055] Step 10.2: Use the Brownian motion of water molecules to traverse the solution space and then update the optimal individual position of the molecules; the mathematical formula for the Brownian motion position update is:

[0056] ;

[0057] ;

[0058] ;

[0059] In the formula, For the Moment The location of each molecule; For A random variable selected from ; For the Snow melting rate of each molecule at the optimal position; for The optimal position of the molecule at the moment; for Suboptimal position of the molecule at the moment; for The third optimal position of the molecule at the moment; for The average position of the molecules at time instant; is a scaling parameter used to control the influence of Brownian motion, and takes the value of (-1, 1); for The average position of the molecules at time instant;

[0060] Step 10.3: Update the position of the molecule according to the snow melting process. The mathematical formula of the snow melting process is:

[0061] ;

[0062] In the formula, is the snow ablation coefficient; is the daily average temperature; is the base temperature;

[0063] The formula is:

[0064] ;

[0065] The mathematical formula for position update is:

[0066] ;

[0067] In the formula, is the snow melt rate; is a random number in the range of [-1, 1]; After the update Moment The location of each molecule; After the update Snow melting rate of each molecule at the optimal position; After update The optimal position of the molecule at the moment; After update The average position of the molecules at time instant; After the update Moment The location of each molecule;

[0068] A maximum number of iterations R is set. When the number of iterations reaches R, the search is stopped and the current best position is regarded as the optimal solution.

[0069] Beneficial technical effects brought by the present invention:

[0070] The present invention can obtain rich information on the motion state and motion environment of the submarine-launched UAV through the various sensors onboard the submarine-launched UAV, and can comprehensively describe the hydrodynamic characteristics of the submarine-launched UAV, providing an accurate data basis for subsequent CFD simulation.

[0071] CFD simulation can be used to obtain parameters that are difficult to obtain with sensors and have a greater impact on the hydrodynamic performance of drones, such as the impact of attached water quality and retained water on the water entry and exit process. Comprehensive water exit status data can make motion evaluation more accurate.

[0072] The convolutional neural network (CNN) can process multi-channel inputs of multivariate time series and effectively capture the spatial relationship between input features. The bidirectional gated recurrent unit (BiGRU) can simultaneously consider past and future information, which helps improve the model's ability to perceive dynamic changes in time series data and is more computationally efficient. The introduction of the self-attention mechanism (Attention) allows the model to more flexibly weight input information at different time steps, and also helps to focus on time points that have a greater impact on the prediction target and process long-term dependencies in the time series, improving the model's ability to perceive the global information of the input sequence when predicting.

[0073] The SAO snow melting hyperparameter optimization method is introduced to optimize the parameters that are difficult to determine in the model. The SAO algorithm itself has advantages such as efficient exploration and utilization strategies and flexible position update equations, which can improve the accuracy of the motion assessment model. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 It is a schematic diagram of a system for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water in the present invention.

[0075] Figure 2 The present invention is a flow chart of a method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water. DETAILED DESCRIPTION

[0076] The specific implementation of the present invention is further described below in conjunction with specific embodiments:

[0077] A system for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water. Figure 1 As shown, it includes a submarine-launched UAV sensing system, a submarine-launched UAV communication system, a data processing and storage system, a CFD simulation analysis system, and a submarine-launched UAV water entry and exit characteristics evaluation platform. The data processing and storage system, the CFD simulation analysis system, and the submarine-launched UAV water entry and exit characteristics evaluation platform are arranged on the mother ship;

[0078] The submarine-launched UAV sensing system includes flow field sensors, Doppler velocimeters, acceleration sensors, temperature, salinity and depth sensors, GPS, and gyroscopes, which collect real-time data on the launch, exit from the water, and ascent of the submarine-launched UAV; the Doppler velocimeter is used to monitor the speed of the submarine-launched UAV, and the acceleration sensor is used to monitor the acceleration of the submarine-launched UAV during its movement; the GPS is used to obtain the position of the submarine-launched UAV; the gyroscope is used to monitor the position data of the submarine-launched UAV; the flow field sensor is used to measure the flow field velocity and pressure; the temperature, salinity and depth sensor is used to provide water temperature, density, salinity and depth.

[0079] The submarine-launched drone communication system is used to upload data to the mother ship for processing and analysis;

[0080] The data processing and storage system is used to process the data from the submarine-launched UAV sensor system and store the processed data in a database; specifically, the data processing and storage system is used for signal noise reduction processing for subsequent analysis, weighted smoothing processing to obtain motion state data without accidental data mutations, extraction mutation processing to obtain motion state data of instantaneous changes in water entry and exit, and simple moving average processing to obtain the most recent stable motion state data of the data stream; feature extraction is performed on the data to obtain time domain feature parameters and frequency feature parameters, and then feature normalization is performed to extract mean, extreme value, and zero-crossing point data.

[0081] The CFD simulation and analysis system is used to obtain data that is difficult for sensors to obtain but has a significant impact on the movement of submarine-launched UAVs. Specifically, the CFD simulation and analysis system is used to create a submarine-launched UAV geometric model based on the data in the data storage system as the initial state parameters, and to perform meshing and mesh independence verification; set flow field boundary conditions, define fluid domains and the position of submarine-launched UAVs; create a simulation physical model of the submarine-launched UAV, and select appropriate fluid models, turbulence models, boundary conditions, and material properties; configure solver settings, including time step, convergence criteria, and number of iterations to meet simulation requirements; perform simulation iterative calculations, and view calculation results through visualization tools, including attached water mass, posture, velocity and other hydrodynamic characteristic parameter data, as well as velocity cloud maps, pressure cloud maps, and surface impact cloud maps during the entry and exit process.

[0082] The Submarine-launched UAV Water Entry and Exit Characteristics Evaluation Platform is used to analyze the hydrodynamic characteristics of the submarine-launched UAV during its entry and exit, and to determine whether the submarine-launched UAV can complete the water entry and exit mission as expected, decide whether the submarine-launched UAV should perform the mission, make corresponding mission adjustments, and remotely control the UAV.

[0083] A method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water. Figure 2 As shown, the above-mentioned system for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water includes the following steps:

[0084] Step 1: Obtain the speed, acceleration, position, posture data of the submarine-launched UAV relative to the water surface depth, flow field speed and density, pressure, water temperature, density, salinity and depth data through the submarine-launched UAV sensor system;

[0085] Step 2: The data collected in step 1 is transmitted to the mother ship through the submarine-launched UAV communication system;

[0086] Step 3: Process the data through the mother ship data processing and storage module, including singular value decomposition denoising, weighted smoothing, extraction mutation processing, and simple moving average processing to obtain data with different characteristics;

[0087] Step 4: Extract features from the processed data, use Fourier transform to obtain time domain feature parameters and frequency feature parameters, normalize the time domain feature parameters and frequency feature parameters, extract mean, extreme value, and zero-crossing point data, and store them in the database;

[0088] Step 5: Determine the initial angle, initial velocity, relative water depth, flow field velocity and density of the UAV relative to the water surface required for the CFD simulation of the submarine-launched UAV, and retrieve the relevant parameters from the database;

[0089] Step 6: Create a simulation physical model of the submarine-launched UAV, perform mesh division and mesh independence verification on the UAV, and select the best simulation network;

[0090] Step 7: Set the flow field boundary conditions and simulation physical model, perform iterative simulation calculations, and obtain the force variation data, velocity variation data, acceleration variation data, angle variation data, flow field evolution data, and corresponding execution results of the submarine-launched UAV entry and exit mission under different conditions. The execution results of the submarine-launched UAV entry and exit mission include successful entry and exit and failure to enter and exit the water;

[0091] Step 8: Divide the data in the database and the data obtained by CFD simulation into training set, validation set and test set;

[0092] Step 9: Construct a CNN-BiGRU-Attention evaluation model, including CNN units, BiGRU units, and Attention units;

[0093] Step 10: Use the training set and validation set to train the CNN-BiGRU-Attention evaluation model, and use the test set to test the trained model to determine whether the model's evaluation ability is optimal. If the model's evaluation ability is not optimal, adjust the relevant hyperparameters;

[0094] Step 11: After optimizing the hyperparameters, continue training the CNN-BiGRU-Attention evaluation model algorithm to obtain a trained CNN-BiGRU-Attention evaluation model;

[0095] Step 12: Evaluate the motion state parameters input by the submarine-launched UAV to determine whether the entry and exit mission can be completed as expected. If the entry and exit mission cannot be completed, control the motion state of the UAV to achieve the conditions for the UAV to enter and exit the water normally.

[0096] Specifically, in step 9, the CNN unit is a convolutional neural network, which effectively extracts features by processing image data and time series data; the CNN unit consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer;

[0097] The mathematical expression of the convolutional layer is:

[0098] ;

[0099] in, For the The convolutional layer Input feature maps or feature data; For the The convolutional layers are located at Weight data of For the Convolutional layer The bias term corresponding to the feature map; As the activation function, the ReLU activation function is selected, and the mathematical expression is:

[0100] ;

[0101] The mathematical expression of the pooling layer is:

[0102] ;

[0103] In the formula, For the The first pooling layer Input feature maps or feature data, For the Pooling layer The weight of a feature map or feature data; is the pooling function; For the The pooling window of the pooling layer, For the Pooling layer The bias term corresponding to the feature map.

[0104] The fully connected layer is used to integrate features for further analysis of global features, and the output layer is connected to the BiGRU unit.

[0105] The BiGRU unit consists of two GRU units with opposite propagation directions, which can process input data in both positive and negative directions, helping to more comprehensively capture time series relationships and process complex time series data. The calculation process of the GRU unit is:

[0106] ;

[0107] ;

[0108] ;

[0109] ;

[0110] In the formula, To reset the gate, To update the gate, is the candidate hidden layer state, is the hidden layer output, is the Sigmoid function, tanh is the activation function, , , are the weight matrices input to the reset gate, update gate, and candidate hidden state, respectively, and are used to map the input vector to the corresponding gate or state; , , are the weight matrices from the previous hidden state to the reset gate, update gate, and candidate hidden state, respectively, which are used to map the hidden state of the previous time step to the corresponding gate or state; is the input vector of the current time step t; The previous time step The hidden state vector of

[0111] The calculation process of BiGRU is:

[0112] ;

[0113] ;

[0114] ;

[0115] In the formula, GRU is the calculation process of the above GRU unit, , are the forward and backward hidden layer outputs, , is the hidden layer output weight, for The hidden layer bias of .

[0116] The Attention unit is an attention mechanism that assigns different weight values ​​to feature data, focuses on key information, and highlights the key features of the data. The mathematical expression is:

[0117] ;

[0118] In the formula, is the query vector, is the key vector, is a value vector.

[0119] In step 10, the SAO algorithm is introduced to optimize the hyperparameters. The hyperparameters are regarded as molecular positions. The SAO algorithm process is as follows:

[0120] Step 10.1: Initialize the algorithm and generate a random population of molecules. The molecular position expression is:

[0121] ;

[0122] In the formula, For the Molecules Dimensional location; , are the upper and lower limits of the molecular position; is a random number in the interval [0,1]; is the number of molecules; is the number of hyperparameters;

[0123] Step 10.2: Use the Brownian motion of water molecules to traverse the solution space and then update the optimal individual position of the molecules; the mathematical formula for the Brownian motion position update is:

[0124] ;

[0125] ;

[0126] ;

[0127] In the formula, For the Moment The location of each molecule; For A random variable selected from ; For the Snow melting rate of each molecule at the optimal position; for The optimal position of the molecule at the moment; for Suboptimal position of the molecule at the moment; for The third optimal position of the molecule at the moment; for The average position of the molecules at time instant; is a scaling parameter used to control the influence of Brownian motion, and takes the value of (-1, 1); for The average position of the molecules at time instant;

[0128] Step 10.3: Update the position of the molecule according to the snow melting process. The mathematical formula of the snow melting process is:

[0129] ;

[0130] In the formula, is the snow ablation coefficient; is the daily average temperature; is the base temperature;

[0131] The formula is:

[0132] ;

[0133] The mathematical formula for position update is:

[0134] ;

[0135] In the formula, is the snow melt rate; is a random number in the range of [-1, 1]; After the update Moment The location of each molecule; After the update Snow melting rate of each molecule at the optimal position; After update The optimal position of the molecule at the moment; After update The average position of the molecules at time instant; After the update Moment The location of each molecule;

[0136] A maximum number of iterations R is set. When the number of iterations reaches R, the search is stopped and the current best position is regarded as the optimal solution.

[0137] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water, characterized in that: A hydrodynamic characteristics evaluation system for a submarine-launched UAV during its entry and exit into water is adopted, comprising a submarine-launched UAV sensing system, a submarine-launched UAV communication system, a data processing and storage system, a CFD simulation analysis system, and a submarine-launched UAV entry and exit characteristics evaluation platform, wherein the data processing and storage system, the CFD simulation analysis system, and the submarine-launched UAV entry and exit characteristics evaluation platform are arranged on a mother ship; The submarine-launched UAV sensing system includes a flow field sensor, a Doppler velocimeter, an acceleration sensor, a temperature-salinity-depth sensor, a GPS, and a gyroscope, which collects data on the launch, exit, and lift-off of the submarine-launched UAV in real time; The submarine-launched UAV communication system is used to upload data to the mother ship for processing and analysis; The data processing and storage system is used to process data from the submarine-launched UAV sensor system and store the processed data in a database; The CFD simulation analysis system is used to obtain data that is difficult for sensors to obtain and has a significant impact on the movement of submarine-launched UAVs; The submarine-launched UAV water entry and exit characteristics evaluation platform is used to analyze the hydrodynamic characteristics of the submarine-launched UAV during the water entry and exit process, and determine whether the submarine-launched UAV can complete the water entry and exit mission as expected, decide whether the submarine-launched UAV performs the mission, make corresponding mission adjustments, and remotely control the UAV; The method comprises the following steps: Step 1: Obtain the speed, acceleration, position, posture data of the submarine-launched UAV relative to the water surface depth, flow field speed and density, pressure, water temperature, density, salinity and depth data through the submarine-launched UAV sensor system; Step 2: The data collected in step 1 is transmitted to the mother ship through the submarine-launched UAV communication system; Step 3: Process the data through the mother ship data processing and storage module, including singular value decomposition denoising, weighted smoothing, extraction mutation processing, and simple moving average processing to obtain data with different characteristics; Step 4: Extract features from the processed data, use Fourier transform to obtain time domain feature parameters and frequency feature parameters, normalize the time domain feature parameters and frequency feature parameters, extract mean, extreme value, and zero-crossing point data, and store them in the database; Step 5: Determine the initial angle of the drone relative to the water surface, initial velocity, relative water surface depth, flow field velocity and density required for the CFD simulation of the submarine-launched drone, retrieve relevant parameters from the database, and obtain the initial angle of the drone relative to the water surface from the posture data; Step 6: Create a simulation physical model of the submarine-launched UAV, perform mesh division and mesh independence verification on the UAV, and select the best simulation network; Step 7: Set the flow field boundary conditions and simulation physical model, perform iterative simulation calculations, and obtain the force variation data, velocity variation data, acceleration variation data, angle variation data, flow field evolution data, and corresponding execution results of the submarine-launched UAV entry and exit mission under different conditions. The execution results of the submarine-launched UAV entry and exit mission include successful entry and exit and failure to enter and exit the water; Step 8: Divide the data in the database and the data obtained by CFD simulation into training set, validation set and test set; Step 9: Construct a CNN-BiGRU-Attention evaluation model, including CNN units, BiGRU units, and Attention units; Step 10: Use the training set and validation set to train the CNN-BiGRU-Attention evaluation model, and use the test set to test the trained model to determine whether the model's evaluation ability is optimal. If the model's evaluation ability is not optimal, adjust the relevant hyperparameters; Step 11: After optimizing the hyperparameters, continue training the CNN-BiGRU-Attention evaluation model algorithm to obtain a trained CNN-BiGRU-Attention evaluation model; Step 12: Evaluate the motion state parameters input by the submarine-launched UAV to determine whether the entry and exit mission can be completed as expected. If the entry and exit mission cannot be completed, control the motion state of the UAV to achieve the conditions for the UAV to enter and exit the water normally.

2. The method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water according to claim 1, characterized in that: The Doppler speedometer is used to monitor the movement speed of the submarine-launched UAV, the acceleration sensor is used to monitor the acceleration of the submarine-launched UAV during its movement; the GPS is used to obtain the position of the submarine-launched UAV; the gyroscope is used to monitor the position data of the submarine-launched UAV; the flow field sensor is used to measure the flow field velocity and pressure; the temperature, salinity and depth sensor is used to provide water temperature, density, salinity and depth.

3. The method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water according to claim 2, characterized in that: The data processing and storage system is used for signal noise reduction processing for subsequent analysis, weighted smoothing processing to obtain motion state data after removing accidental data mutations, extraction mutation processing to obtain motion state data of instantaneous changes in water entry and exit, and simple moving average processing to obtain the most recent stable motion state data of the data stream; feature extraction is performed on the data to obtain time domain feature parameters and frequency feature parameters, and then feature normalization is performed to extract mean, extreme value, and zero-crossing point data.

4. The method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water according to claim 3, characterized in that: The CFD simulation analysis system is used to create a geometric model of a submarine-launched UAV based on data in a data storage system as initial state parameters, and to perform mesh division and mesh independence verification; Set flow field boundary conditions, define fluid domain and submarine-launched UAV position; Create a simulation physical model of the submarine-launched UAV and select appropriate fluid models, turbulence models, boundary conditions, and material properties; Configure the solver settings, including time step, convergence criterion, and number of iterations, to meet the simulation requirements; perform simulation iterative calculations, and view the calculation results through visualization tools, including hydrodynamic characteristic parameter data such as attached water mass, posture, and velocity, as well as velocity cloud maps, pressure cloud maps, and water surface impact cloud maps during the entry and exit process.

5. The method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water according to claim 4, characterized in that: The CNN unit consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer; The mathematical expression of the convolutional layer is: ； in, For the The convolutional layer Input feature maps or feature data; For the The convolutional layers are located at Weight data of For the Convolutional layer The bias term corresponding to the feature map; As the activation function, the ReLU activation function is selected, and the mathematical expression is: ； The mathematical expression of the pooling layer is: ； In the formula, For the The first pooling layer Input feature maps or feature data, For the Pooling layer The weight of a feature map or feature data; is the pooling function; For the The pooling window of the pooling layer, For the Pooling layer The bias term corresponding to the feature map.

6. The method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water according to claim 5, characterized in that: The BiGRU unit includes two GRU units with opposite propagation directions. The calculation process of the GRU unit is: ； ； ； ； In the formula, To reset the gate, To update the gate, is the candidate hidden layer state, is the hidden layer output, is the Sigmoid function, tanh is the activation function, , , are the weight matrices input to the reset gate, update gate, and candidate hidden state, respectively, and are used to map the input vector to the corresponding gate or state; , , are the weight matrices from the previous hidden state to the reset gate, update gate, and candidate hidden state, respectively, which are used to map the hidden state of the previous time step to the corresponding gate or state; is the input vector of the current time step t; The previous time step The hidden state vector of The calculation process of BiGRU is: ； ； ； In the formula, GRU is the calculation process of the above GRU unit, , are the forward and backward hidden layer outputs, , is the hidden layer output weight, for The hidden layer bias of .

7. The method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water according to claim 6, characterized in that: The Attention unit is an attention mechanism that assigns different weight values ​​to feature data, focuses on key information, and highlights the key features of the data. The mathematical expression is: ； In the formula, is the query vector, is the key vector, is a value vector.

8. The method for evaluating the hydrodynamic characteristics of a submarine-launched UAV during its entry and exit into water according to claim 7, characterized in that: In step 10, the SAO algorithm is introduced to optimize the hyperparameters, and the hyperparameters are regarded as the molecular positions. The SAO algorithm process is as follows: Step 10.1: Initialize the algorithm and generate a random population of molecules. The molecular position expression is: ； In the formula, For the Molecules Dimensional location; , are the upper and lower limits of the molecular position; is a random number in the interval [0,1]; is the number of molecules; is the number of hyperparameters; Step 10.2: Use the Brownian motion of water molecules to traverse the solution space and then update the optimal individual position of the molecules; the mathematical formula for the Brownian motion position update is: ； ； ； In the formula, For the Moment The location of each molecule; For A random variable selected from ; For the Snow melting rate of each molecule at the optimal position; for The optimal position of the molecule at the moment; for Suboptimal position of the molecule at the moment; for The third optimal position of the molecule at the moment; for The average position of the molecules at time instant; is a scaling parameter used to control the influence of Brownian motion, and takes the value of (-1, 1); for The average position of the molecules at time instant; Step 10.3: Update the position of the molecule according to the snow melting process. The mathematical formula of the snow melting process is: ； In the formula, is the snow ablation coefficient; is the daily average temperature; is the base temperature; The formula is: ； The mathematical formula for position update is: ； In the formula, is the snow melt rate; is a random number in the range of [-1, 1]; After the update Moment The location of each molecule; After the update Snow melting rate of each molecule at the optimal position; After update The optimal position of the molecule at the moment; After update The average position of the molecules at time instant; After the update Moment The location of each molecule; A maximum number of iterations R is set. When the number of iterations reaches R, the search is stopped and the current best position is regarded as the optimal solution.

Images