Underwater thruster control method and system based on tensor identification and fuzzy control

Abstract

The present invention provides an underwater thruster control method and system based on tensor identification and fuzzy control, which is suitable for improving propulsion efficiency and adjustment intelligence under complex flow fields. The method constructs a wake simulation model based on the thruster geometric parameters and boundary conditions, derives the rotation tensor and strain tensor after acquiring the flow field data, and extracts the tensor field index through function space mapping to determine the target grid area. The high-curl candidate area is screened by the spectral clustering algorithm, the vortex structure is identified and its multi-dimensional features such as scale, intensity, axial direction and vortex core position are extracted. The state vector is constructed in combination with the current propulsion efficiency and flow field disturbance parameters, and the adaptive fuzzy neural network model is input to infer the relationship between the vortex and the operating state, and the thruster speed and angle of attack adjustment amount are output to achieve intelligent response to the wake disturbance and energy efficiency optimization.

Description

Technical Field

[0001] The present application relates to the technical field of underwater thrusters, and in particular to an underwater thruster control method and system based on tensor identification and fuzzy control. Background Art

[0002] As the primary power output unit for underwater platforms like submersibles, underwater propulsion systems (EPs) have a significant impact on their performance in terms of endurance, stability, and adaptability to complex operating conditions. Most existing propulsion systems still rely heavily on static geometry and linear speed control in their design and control strategies, lacking a deep integration of fluid disturbance response mechanisms in variable operating environments.

[0003] During actual propulsion operation, the complex interaction between the high-speed rotation of the blades and the water often forms a series of dynamically changing vortex structures in the wake flow field. These vortices not only disrupt the distribution of local propulsion force but can also induce instabilities such as cavitation and turbulence enhancement, disrupting the energy transfer path and significantly affecting propulsion efficiency and system stability.

[0004] However, traditional control methods typically only capture macroscopic parameters such as the propeller's rotational speed, target speed, or water depth, lacking a mechanism for identifying and utilizing the vorticity field and disturbance structure within the wake region. This "lack of perception" prevents the control system from effectively responding to the evolution of the flow pattern in the wake, leading to problems such as propulsion performance degradation and system response hysteresis under high load or variable flow conditions. This has become a key bottleneck restricting the further advancement of underwater intelligent propulsion systems. Summary of the Invention

[0005] In order to solve the above technical problems, the present application provides an underwater thruster control method and system based on tensor identification and fuzzy control.

[0006] The technical solution provided in this application is described below:

[0007] The first aspect of the present application provides an underwater thruster control method based on tensor identification and fuzzy control, comprising:

[0008] Based on the propeller's geometric parameters and operating boundary conditions, a numerical simulation model of the wake region is constructed, and simulations are performed to obtain basic flow field data.

[0009] Performing tensor field derivation processing based on the flow field basic data to obtain a rotation tensor and a strain tensor describing rotation and deformation characteristics;

[0010] Mapping the rotation tensor and the strain tensor to a first function space and a second function space, respectively, to obtain a local tensor field index group, and determining a target grid area based on the local tensor field index group;

[0011] generating a curl feature according to the rotation tensor, and constructing a graph structure model based on the curl feature to perform spectral clustering to obtain a candidate grid area, wherein the curl feature of the candidate grid area reaches a preset threshold;

[0012] performing aggregation processing on each grid unit in the candidate grid area and determining a target vortex area, wherein the target vortex area has rotation dominance in the candidate grid area;

[0013] For the identified target vortex region, extract multi-dimensional vortex structure characteristic data including at least vortex scale, vortex intensity, rotation axis and vortex core position;

[0014] Collecting current propulsion efficiency indicators and flow field disturbance parameters, and constructing a multidimensional state vector in combination with the multidimensional vortex structure characteristic data;

[0015] The multidimensional state vector is input into a preconfigured adaptive fuzzy neural network model to dynamically infer the correlation between the vortex intensity and the propeller operating state, and the target speed adjustment amount and blade angle adjustment amount for adjusting the propeller operating state are output.

[0016] Optionally, generating a curl feature according to the rotation tensor, and constructing a graph structure model based on the curl feature to perform spectral clustering to obtain a candidate grid area, wherein the curl feature of the candidate grid area reaches a preset threshold, includes:

[0017] Calculating a curl feature of each grid cell based on the rotation tensor, wherein the curl feature is used to characterize the rotation intensity of the local flow field;

[0018] The grid cells containing the curl features are constructed into a graph structure model, in which nodes correspond to grid cells and edge weights are set based on the difference in curl features between adjacent nodes.

[0019] Applying a spectral clustering algorithm to the graph structure model and performing unsupervised partitioning on the grid cells to obtain clustering results;

[0020] Clusters whose curl characteristics exceed a preset threshold are screened out from the clustering results and determined as candidate grid areas.

[0021] Optionally, mapping the rotation tensor and the strain tensor to the first function space and the second function space respectively to obtain a local tensor field index group, and determining the target grid area based on the local tensor field index group includes:

[0022] Mapping the rotation tensor to a first function space based on rotation characteristics, where the first function space is a characteristic mode domain constructed based on tensor eigenvalue decomposition, and is used to reflect local vortex distribution characteristics;

[0023] Mapping the strain tensor to a second function space dominated by shear deformation and principal stress directions, wherein the second function space is used to represent the anisotropic behavior of local deformation;

[0024] Extracting a first tensor response eigenvalue related to the direction of the vortex principal axis in the first function space, and extracting a second tensor response eigenvalue related to the direction of the maximum principal strain in the second function space;

[0025] constructing a local tensor field index group based on the first tensor response eigenvalue and the second tensor response eigenvalue, for characterizing the rotation dominance and shear coupling characteristics of the grid area;

[0026] A target grid region is determined based on the local tensor field indicator group.

[0027] Optionally, determining the target grid area based on the local tensor field indicator group includes:

[0028] For each grid cell in the wake region, the corresponding first tensor response eigenvalue and second tensor response eigenvalue are obtained respectively;

[0029] Based on a preset functional relationship, combining the first tensor response eigenvalue and the second tensor response eigenvalue to obtain a combined eigenvalue of each grid cell;

[0030] Screening out grid cells that do not meet the rotation dominance condition according to the relationship between the combined eigenvalue and the first threshold;

[0031] For the retained mesh elements, calculate the angle information between the main direction of the rotation tensor and the main direction of the strain tensor;

[0032] A target grid area that meets the rotation dominance condition is screened out according to the angle information.

[0033] Optionally, performing tensor field derivation processing based on the flow field basic data to obtain a rotation tensor and a strain tensor describing rotation and deformation characteristics includes:

[0034] Calculating a velocity gradient tensor for each grid cell based on a velocity vector in the flow field basic data;

[0035] Decomposing the velocity gradient tensor into an antisymmetric part and a symmetric part, wherein the antisymmetric part corresponds to a rotation tensor and is used to describe the rotation characteristics of the local fluid, and the symmetric part corresponds to a strain tensor and describes the deformation characteristics of the local fluid;

[0036] The eigenvalues ​​and eigenvectors of the rotation tensor and strain tensor are calculated respectively.

[0037] Optionally, the condition for satisfying rotation dominance includes that the combined eigenvalue is greater than or equal to the first threshold.

[0038] Optionally, performing aggregation processing on each grid unit in the candidate grid area and determining a target vortex area, wherein the target vortex area has rotation dominance in the candidate grid area, comprises:

[0039] According to the spatial proximity relationship, multiple adjacent grid cells that meet the rotation dominance condition are divided into a connected region as a candidate vortex region;

[0040] Calculating a comprehensive vortex strength index within each candidate vortex region, wherein the comprehensive vortex strength index includes vortex core position coordinates, vortex strength, vortex core scale, and local energy consumption density;

[0041] According to the preset vortex intensity threshold and vortex core size threshold, the target vortex area that meets the conditions of vortex intensity threshold and rotation dominance is screened out.

[0042] Optionally, the step of inputting the multidimensional state vector into a preconfigured adaptive fuzzy neural network model to dynamically infer the correlation between the vortex intensity and the propeller operating state, and outputting a target speed adjustment value and a blade angle of attack adjustment value for adjusting the propeller operating state includes:

[0043] Fuzzifying each variable in the multidimensional state vector and mapping it to a corresponding fuzzy set to form a fuzzy input variable;

[0044] According to a preset fuzzy rule base, determining the activation strength of each fuzzy rule in the fuzzy rule base based on the fuzzy input variable;

[0045] Determine a target fuzzy rule based on the activation strength, and perform comprehensive simulation on the target fuzzy rule through a fuzzy simulation module in an adaptive fuzzy neural network model to obtain an output fuzzy set;

[0046] The output fuzzy set is defuzzified to obtain a numerical target speed adjustment amount and a blade angle of attack adjustment amount.

[0047] A second aspect of the present application provides an underwater thruster control system based on tensor identification and fuzzy control, characterized by comprising:

[0048] A simulation unit is used to construct a numerical simulation model of the wake area based on the geometric parameters and operating boundary conditions of the propeller, and perform simulation to obtain basic flow field data;

[0049] A tensor field unit is used to perform tensor field derivation processing based on the flow field basic data to obtain a rotation tensor and a strain tensor describing rotation and deformation characteristics;

[0050] a function mapping unit, configured to map the rotation tensor and the strain tensor to a first function space and a second function space, respectively, to obtain a local tensor field index group, and to determine a target grid area based on the local tensor field index group;

[0051] a pre-screening unit, configured to generate a curl feature according to the rotation tensor, and construct a graph structure model based on the curl feature to perform spectral clustering to obtain a candidate grid area, wherein the curl feature of the candidate grid area reaches a preset threshold;

[0052] a grid unit aggregation unit for performing aggregation processing on each grid unit in the candidate grid area and determining a target vortex area, wherein the target vortex area has rotation dominance in the candidate grid area;

[0053] A data extraction unit is used to extract multi-dimensional vortex structure characteristic data including at least vortex scale, vortex intensity, rotation axis and vortex core position from the identified target vortex area;

[0054] A data acquisition unit, configured to collect current propulsion efficiency indicators and flow field disturbance parameters, and construct a multidimensional state vector based on the multidimensional vortex structure characteristic data;

[0055] The multidimensional state vector is input into a preconfigured adaptive fuzzy neural network model to dynamically infer the correlation between the vortex intensity and the propeller operating state, and the target speed adjustment amount and blade angle adjustment amount for adjusting the propeller operating state are output.

[0056] A third aspect of the present application provides an underwater thruster control system based on tensor identification and fuzzy control, the device comprising:

[0057] processor, memory, input and output units, and buses;

[0058] The processor is connected to the memory, the input and output unit, and the bus;

[0059] The memory stores a program, and the processor calls the program to execute the first aspect and any optional method in the first aspect.

[0060] It can be seen from the above technical solutions that this application has the following advantages:

[0061] By mapping the rotation tensor and strain tensor to different function spaces respectively, the present invention achieves the effective separation and high-dimensional expression of the rotation characteristics and deformation characteristics in the flow field, and enhances the accuracy and robustness of vortex structure identification. By combining the local tensor field indicators in the two function spaces to identify the target grid area, it is possible to more finely screen out the rotation-dominated vortex area, avoiding the misjudgment problem caused by shear deformation in traditional methods, thereby improving the accuracy and reliability of vortex extraction. This step provides a more accurate and stable input data basis for subsequent dynamic reasoning based on fuzzy neural networks, further ensuring the effect of propeller intelligent control and the sensitivity of system response.

[0062] The present invention constructs a numerical simulation model of the wake region based on the propeller geometric parameters and operating boundary conditions, and combines the tensor field analysis of the rotation tensor and the strain tensor to achieve accurate identification and positioning of the vortex structure in the flow field. The multidimensional state vector is used to comprehensively consider the target multidimensional vortex structure characteristic data, propulsion efficiency index and flow field disturbance parameters, and combined with the pre-configured adaptive fuzzy neural network model, the correlation between the vortex intensity and the propeller operating state is dynamically inferred, and the accurate target speed adjustment amount and blade angle of attack adjustment amount are effectively output. The method can intelligently respond to flow field disturbances, realize real-time optimization and adjustment of the propeller operating state, significantly improve the operating efficiency and stability of the underwater propeller, reduce energy consumption and mechanical wear, and improve the overall performance and life of the underwater propulsion system. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in this application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0064] Figure 1 This is a flow chart of an embodiment of the underwater thruster control method based on tensor identification and fuzzy control provided in this application;

[0065] Figure 2 This is a flowchart of a specific embodiment of step S102 in the underwater thruster control method based on tensor identification and fuzzy control provided in this application;

[0066] Figure 3 This is a flowchart of a specific embodiment of step S103 in the underwater thruster control method based on tensor identification and fuzzy control provided in this application;

[0067] Figure 4This is a flowchart of a specific embodiment of step S105 in the underwater thruster control method based on tensor identification and fuzzy control provided in this application;

[0068] Figure 5 This is a flowchart of a specific embodiment of step S106 in the underwater thruster control method based on tensor identification and fuzzy control provided in this application;

[0069] Figure 6 This is a schematic structural diagram of an embodiment of an underwater thruster control system based on tensor identification and fuzzy control provided in this application;

[0070] Figure 7 This is a schematic structural diagram of another embodiment of an underwater thruster control system based on tensor identification and fuzzy control provided in this application. DETAILED DESCRIPTION

[0071] In this application, in order to achieve accurate identification and dynamic control of vortex structures in the flow field of underwater propellers, the definition and use of multiple physical parameters and mathematical quantities are involved. The relevant terms and setting standards are explained as follows:

[0072] Vorticity is the rotation rate of velocity in a vector field relative to space, and is defined as the curl of the velocity vector field. In this application, the curl modulus is used to represent the local rotation intensity in the flow field and is an important basis for determining whether a vortex structure exists. To facilitate engineering implementation, the curl threshold is usually set based on the statistical distribution of simulation samples, for example, by adding the mean value and standard deviation to determine the identification threshold, with a typical value range of 0.1 to 1.5 S. -1 between.

[0073] Tensor response features refer to the response quantities obtained by tensorizing velocity fields, pressure fields, and so on. Common examples include velocity gradient tensors and stress tensors. In this paper, two function spaces are introduced to map tensor features: the original tensor space, which retains the complete second-order tensor information for each measurement point; and the combined feature space, which uses eigenvalue functions (such as the sum of squared eigenvalues) to map tensors into a single scalar or vector form for simplified processing and cluster analysis.

[0074] Spectral clustering is used to classify vortex features and identify typical vortex patterns in a flow field. Based on a similarity matrix, this method constructs a graph Laplacian matrix and extracts the top k eigenvectors to achieve dimensionality reduction and clustering. Similarity is typically calculated using a Gaussian kernel function.

[0075] Vortex features include but are not limited to: vortex core location (curl or Q-criterion extreme point), rotation intensity, principal axis direction (tensor principal eigenvector), and scale (feature region envelope volume or projected diameter). Vortex features can be extracted using fluid simulation software (such as Fluent or OpenFOAM) combined with post-processing tools (such as ParaView) and used as input for subsequent control model training.

[0076] A fuzzy neural network (FNN), a control model that integrates fuzzy rules and neural architecture, is used for the propeller dynamic control in this invention. The network input is the extracted multidimensional vortex structure features. The fuzzy layer fuzzifies the input using Gaussian or triangular membership functions. The rule layer outputs a response based on preset or self-learned rules, ultimately controlling the propeller's speed and angle of attack. The network structure can be a five-layer structure: input layer - fuzzy layer - rule layer - normalization layer - output layer.

[0077] The control variables in the present invention are output variables of the control system, mainly including the propeller speed change (unit: rpm) and the blade angle of attack change (unit: °). The control variables are output in the form of increments.

[0078] A tensor is a mathematical object that can represent multidimensional data relationships. It is a natural extension of scalars (0th-order tensors) and vectors (1st-order tensors). In fluid mechanics, common tensors include the velocity gradient tensor (VGT), stress tensor, and strain rate tensor. For example, the velocity gradient tensor is defined as the matrix of partial derivatives of the velocity field with respect to spatial coordinates.

[0079] A tensor field is a physical field in which each spatial point is associated with a tensor, a further extension of a vector field. For example, a velocity gradient tensor field is a collection of velocity gradient tensors at every point in a three-dimensional flow field. It contains rich flow structure information and can be used to identify vortices and flow anomalies.

[0080] Vortex scale is used to characterize the spatial size of a vortex region. It can be estimated by envelope diameter, area, or distance between characteristic points. For example, it is defined as the maximum circumscribed sphere radius of a region with a curl greater than a threshold.

[0081] Propeller geometric parameters, such as pitch ratio, blade twist angle, blade root / tip thickness ratio, inlet angle, etc., have a direct impact on the flow field disturbance characteristics.

[0082] See also Figure 1 This application first provides an embodiment of an underwater thruster control method based on tensor identification and fuzzy control, which includes:

[0083] S101. Based on the geometric parameters and operating boundary conditions of the propeller, a numerical simulation model of the wake region is constructed, and simulation is performed to obtain basic flow field data.

[0084] First, a numerical simulation model of the wake region is constructed based on the geometric parameters of the underwater propeller to be controlled (including propeller diameter, pitch, number of blades, etc.) and its operating boundary conditions (such as propulsion speed, fluid density, viscosity coefficient, boundary inlet velocity distribution, and outlet pressure). This simulation model uses a three-dimensional incompressible flow field model, and a sliding mesh technique is introduced in the numerical solution to accommodate dynamic mesh updates during propeller rotation.

[0085] After performing numerical simulations, we obtain basic three-dimensional flow field data in the wake region, including the velocity vector field (Ux, Uy, Uz) and the pressure scalar field (P). This data serves as input for subsequent tensor processing and structural identification.

[0086] S102, performing tensor field derivation processing based on the flow field basic data to obtain a rotation tensor and a strain tensor describing rotation and deformation characteristics;

[0087] Based on the velocity vector field data obtained above, a velocity gradient tensor is calculated for each grid cell. This tensor describes the velocity variation trend at that point in all directions. Furthermore, based on the velocity gradient tensor, two basic descriptions are obtained through tensor transformation: the rotation tensor Ω and the strain tensor S.

[0088] Among them, the rotation tensor Ω represents the rotation of the local fluid and is generally defined as the antisymmetric part of the velocity gradient tensor, while the strain tensor S is the symmetric part of the velocity gradient tensor and is used to describe the degree of deformation of the fluid element.

[0089] For details, see Figure 2 , an optional implementation of this step includes:

[0090] S1021. Calculate the velocity gradient tensor of each grid cell based on the velocity vector in the basic flow field data;

[0091] Based on the obtained three-dimensional velocity vector field V=(u, v, w), the velocity gradient tensor is calculated for each grid cell. , which is a 3×3 matrix:

[0092] ;

[0093] The above partial derivatives are obtained by calculating the velocity data on the simulation grid using the central difference method or other high-order finite difference schemes.

[0094] S1022. Decomposing the velocity gradient tensor into an antisymmetric part and a symmetric part, wherein the antisymmetric part corresponds to a rotation tensor and is used to describe the rotation characteristics of the local fluid, and the symmetric part corresponds to a strain tensor and is used to describe the deformation characteristics of the local fluid;

[0095] The local velocity gradient tensor Decompose into symmetric part and antisymmetric part, that is:

[0096] The strain tensor S is the symmetric part of the velocity gradient tensor and is expressed as:

[0097] ;

[0098] The rotation tensor Ω is the antisymmetric part of the velocity gradient tensor and is expressed as:

[0099]

[0100] in, The strain tensor S is used to reflect the tensile and compressive deformation of the local fluid particles, while the rotation tensor Ω describes the rotational motion characteristics of the local fluid.

[0101] S1023. Calculate the eigenvalues ​​and eigenvectors of the rotation tensor and strain tensor respectively.

[0102] The rotation tensor Ω and the strain tensor S are subjected to eigenvalue decomposition respectively, and their eigenvalues ​​and corresponding eigenvectors are calculated.

[0103] Specifically, for each tensor T, solve its characteristic equation:

[0104] det(T-λI)=0;

[0105] Where λ is the eigenvalue of the tensor, I is the identity matrix, and det represents the determinant. This equation yields three eigenvalues, λ1, λ2, and λ3, which are used to analyze the rotational dominance and deformation trends of the local fluid structure. These eigenvalues ​​serve as the input for constructing the local tensor field index set and function space mapping.

[0106] S103, mapping the rotation tensor and the strain tensor to the first function space and the second function space respectively, obtaining a local tensor field index group, and determining a target grid area based on the local tensor field index group;

[0107] The rotation tensor and strain tensor are mapped to two different function spaces, namely the first function space and the second function space, respectively. The eigenvalue decomposition is used to extract the tensor features and construct a local tensor field index group. For example, the following method can be used:

[0108] The tensor obtained by summing S and Ω is subjected to eigenvalue analysis to obtain an eigenvalue set;

[0109] In an alternative embodiment, see Figure 3 , the implementation of step S103 includes:

[0110] S1031, mapping the rotation tensor to a first function space based on rotation characteristics, where the first function space is a characteristic mode domain constructed based on tensor eigenvalue decomposition, and is used to reflect local vortex distribution characteristics;

[0111] The rotation tensor Ω obtained in step S102 is mapped to a first function space based on rotational features. This first function space is an eigenmode domain constructed based on tensor eigenvalue decomposition. It is essentially a structural representation of the tensor eigenvalues ​​and the directions of their corresponding eigenvectors. It can be used to characterize the directionality, rotation intensity, and rotation continuity of the local fluid rotation axis.

[0112] In this embodiment, the response index under the function space is defined as:

[0113] ;

[0114] in, are the three eigenvalues ​​of the rotation tensor, Indicates the main response intensity in the direction of the local vortex axis.

[0115] S1032. Mapping the strain tensor to a second function space based on shear deformation and principal stress directions, where the second function space is used to represent anisotropic behavior of local deformation;

[0116] The strain tensor S in step S102 is mapped to a second function space based on shear deformation and principal stress directions. This second function space is used to identify the anisotropy of fluid deformation in different directions, reflecting structural behaviors such as the principal directions of local tension and compression and the degree of shear strain.

[0117] The response index in this space can be defined as:

[0118] ;

[0119] in is the eigenvalue of the strain tensor, Represents the tensor strength along the principal strain directions. The principal strain axis features constructed in the second function space can be used to assist in determining whether high shear regions exist and whether these regions are coupled to rotational behavior.

[0120] S1033. Extracting a first tensor response eigenvalue related to the direction of the vortex principal axis in the first function space, and extracting a second tensor response eigenvalue related to the direction of the maximum principal strain in the second function space;

[0121] Extract the first tensor response eigenvalue related to the vortex main axis direction in the first function space , and extract the second tensor response eigenvalue related to the maximum principal strain direction in the second function space , as a characterization of local tensor behavior.

[0122] In order to enhance the robustness of structure determination, the rotation dominant ratio index can be defined , expressed as:

[0123] ;

[0124] in, A small constant to prevent the denominator from being zero. This metric is used to assess how dominant the rotational behavior is in the local tensor response.

[0125] S1034. Constructing a local tensor field index group based on the first tensor response eigenvalue and the second tensor response eigenvalue, for characterizing the rotational dominance and shear coupling characteristics of the grid area;

[0126] Based on the above extraction 、 、 A local tensor field index group is constructed, which can characterize the multidimensional properties of each grid cell in terms of rotational dominance and shear coupling behavior.

[0127] S1035. Determine a target grid area based on the local tensor field indicator group.

[0128] Perform index traversal analysis on the entire simulation grid domain and select grid cells that meet the following constraints as the target grid area:

[0129] > : The local rotation intensity is higher than the vortex detection threshold;

[0130] > : Rotation dominates the behavior of tensors;

[0131] Optional: < .

[0132] in, 、 、 The grid cells that meet the above conditions will constitute the rotation dominant area and enter the subsequent spatial aggregation processing steps S104 and S105.

[0133] In another optional embodiment, the determination of rotational dominance may be accomplished by using the Q-criterion or the λ2 criterion;

[0134] In the pre-screening stage, the Q criterion is used to quickly exclude shear-dominated areas and preliminarily screen areas that meet the Q>0 requirement;

[0135] In the fine judgment stage, the λ2 criterion is used to further extract grid cells that meet λ2<0 in the area that meets Q>0, forming a target grid area that meets the rotation-dominant trend.

[0136] It is worth noting that both the first and second function spaces described in step S103 are derived from tensor decomposition of the velocity gradient tensor. However, unlike traditional vortex criteria, which are used only for structural identification, this embodiment constructs a more parameterized and responsive function space and tensor field index set. This allows the subsequent intelligent control model to input richer, continuous, and coupled information, enhancing the control process's ability to analyze and respond to flow field states.

[0137] S104, generating a curl feature according to the rotation tensor, and constructing a graph structure model based on the curl feature to perform spectral clustering to obtain a candidate grid area, wherein the curl feature of the candidate grid area reaches a preset threshold;

[0138] In this embodiment, step S104 involves generating a curl feature for the rotation tensor and constructing a graph structure model based on the curl feature to perform spectral clustering, thereby obtaining candidate grid regions with high curl features. First, based on the obtained rotation tensor, the curl feature of each grid cell is calculated. All grid cells are treated as nodes in a graph structure. Edges are established between nodes based on physical adjacency or fluid similarity, and the weights of the edges in the graph are calculated based on the difference in curl.

[0139] Based on the constructed graph structure model, a spectral clustering algorithm is used to perform cluster analysis on the grid units, which specifically includes the steps of solving the graph Laplacian matrix, extracting eigenvectors and K-means clustering. Spectral clustering can aggregate spatially discontinuous grid units with similar curls into regions with consistent features while maintaining local curl consistency. After clustering is completed, the average curl characteristics of each cluster region are counted, and the clustering results with an average curl exceeding a preset threshold are screened as candidate grid regions. These candidate grid regions have obvious rotation dominance and are key objects for subsequent vortex structure identification and dynamic control reasoning. Through this step, the system can effectively extract representative rotation regions in complex flow fields, thereby improving the efficiency and accuracy of subsequent processing.

[0140] Specifically, this application provides an implementation of step S104, which specifically includes:

[0141] The curl feature of each grid cell is calculated based on the rotation tensor, and the curl feature is used to characterize the rotation intensity of the local flow field. The grid cells containing the curl feature are constructed into a graph structure model, in which nodes in the graph structure model correspond to grid cells, and edge weights are set based on the difference in curl features between adjacent nodes. A spectral clustering algorithm is applied to the graph structure model, and the grid cells are partitioned in an unsupervised manner to obtain clustering results. Cluster clusters whose curl features exceed a preset threshold are screened out from the clustering results and determined as candidate grid areas.

[0142] In this implementation, the curl feature of each grid cell is calculated based on the rotation tensor, and the curl feature is used to characterize the rotation intensity in the local flow field; the grid cells containing the curl feature are constructed into a graph structure model, wherein the nodes in the graph structure correspond to each grid cell, and the weights of the edges are set according to the difference in the curl features between adjacent nodes to reflect the similarity between the nodes; a spectral clustering algorithm is applied to the graph structure model, and the grid cells are divided using unsupervised learning to obtain multiple clustering results; finally, cluster clusters whose curl features exceed a preset threshold are screened out from the clustering results, and these cluster clusters are determined to be candidate grid areas with significant rotation features for subsequent vortex identification and control strategy formulation.

[0143] S105. Aggregate the grid cells in the candidate grid area and determine a target vortex area, where the target vortex area is rotationally dominant in the candidate grid area; perform spatial aggregation on the target grid area, grouping adjacent continuous rotationally dominant grid cells into a vortex cluster to form one or more closed or semi-closed vortex areas.

[0144] The corresponding structural parameters of each vortex cluster region are further extracted to form the target multi-dimensional vortex structural characteristic data. This structural data may include but is not limited to: vortex center position, radius, vortex axis direction; vortex internal rotation intensity, axial velocity distribution; local pressure gradient and velocity vorticity, etc.

[0145] S106, extracting multi-dimensional vortex structure characteristic data including at least vortex scale, vortex intensity, rotation axis and vortex core position for the identified target vortex region;

[0146] The data extracted from the target vortex structure is combined with the propulsion efficiency indicators in the current propeller operating state (such as propulsion distance per unit power consumption) and the flow field disturbance parameters obtained by detection (such as incoming flow changes and wake fluctuation frequency) to jointly construct a multi-dimensional state vector.

[0147] This input vector, as the input dimension of the fuzzy neural network, can dynamically reflect the coupling relationship between the vortex structure and the propulsion environment.

[0148] For details, see Figure 4 , an optional implementation includes:

[0149] S1061, collecting target multi-dimensional vortex structure characteristic data, wherein the target multi-dimensional vortex structure characteristic data includes vortex intensity, vortex core position, vortex core size, and local energy consumption density;

[0150] For each target vortex region extracted in step S105, the corresponding key structural parameters are obtained to form a target multi-dimensional vortex structure feature dataset. This data includes but is not limited to the following indicators:

[0151] Vortex strength ωs: can be calculated from the vorticity modulus, axial velocity gradient or local rotation tensor norm, and is used to characterize the rotational energy level of the vortex;

[0152] Vortex core position Xc: the spatial coordinate point corresponding to the dominant eigenvalue of rotation in the vortex cluster (such as the minimum value of λ2<0);

[0153] Vortex core scale Rc: represents the local size range of the vortex structure, which can be estimated by the maximum response radius or characteristic length;

[0154] Local energy dissipation density Ed: Calculated based on the strain tensor modulus and shear power dissipation function, reflecting the degree of energy dissipation in the area.

[0155] S1062. Collect current propulsion efficiency indicators and flow field disturbance parameters;

[0156] Synchronously collect the current thruster working status parameters, including but not limited to:

[0157] Unit propulsion efficiency η = PVf, where Vf is the propulsion speed and P is the power consumption per unit time;

[0158] Wake disturbance frequency fw: obtained by spectral analysis of the main frequency component of velocity fluctuations in the flow field;

[0159] Inflow disturbance angle θin: records the deviation angle of the inflow direction, reflecting the impact of attitude changes on the flow field structure;

[0160] Fluid Reynolds number Re: calculated from propulsion velocity, characteristic length and fluid viscosity parameters to unify dimensional effects.

[0161] S1063, normalizing the collected data;

[0162] The vortex structure parameters, propulsion efficiency index and disturbance parameters collected above are normalized respectively to eliminate the coupling interference between different dimensions and physical dimensions. The optional methods include min-max linear normalization or Z-score normalization.

[0163] S1064. Combining the normalized target multi-dimensional vortex structure characteristic data, propulsion efficiency index, and flow field disturbance parameter into a multi-dimensional state vector.

[0164] The normalized target multi-dimensional vortex structure characteristic data {ωs, Xc, Rc, Ed}, propulsion efficiency index η and flow field disturbance parameters are:

[0165] {fw, θin, Re};

[0166] According to the preset dimensions, the combination is used as a whole vector input to form a multi-dimensional state vector S_input for the intelligent reasoning model:

[0167] S_input=[ωs, Rc, Ed, η, fw, θin, Re].

[0168] Optionally, the characteristic change rate of the historical frame (such as the change rate of vortex intensity) or the change rate of propulsion efficiency is introduced into the input vector to enhance the model's perception of time dynamic trends.

[0169] This input vector is used for the fuzzy neural network dynamic reasoning in the subsequent step S106, which can reflect the coupling state between the vortex structure and the propulsion environment in real time, thereby driving the intelligent adjustment of the propeller operating parameters.

[0170] S107, collecting current propulsion efficiency indicators and flow field disturbance parameters, and constructing a multidimensional state vector in combination with the multidimensional vortex structure characteristic data;

[0171] In this embodiment, in order to realize intelligent regulation and control of the propeller operating state, after completing the vortex area identification and its structural feature extraction, the system further collects multiple parameters including the current operating efficiency of the propeller and the wake disturbance characteristics, and integrates them with the multi-dimensional vortex structure feature data to construct a multi-dimensional state vector for intelligent model input.

[0172] Specifically, the steps include:

[0173] Propulsion efficiency index collection: The system obtains the propulsion efficiency data of the current propeller under the target operating conditions in real time. The efficiency can be expressed as the ratio of the net thrust output work per unit time to the system power consumption. It can also include operating condition-related efficiency parameters such as blade propulsion efficiency and forward speed per unit input power consumption.

[0174] Acquisition of flow field disturbance parameters: By deploying pressure sensors and flow velocity monitoring modules in the wake area or through simulation model deduction, disturbance characteristic quantities such as local flow velocity fluctuations, pressure gradient changes, Reynolds number distribution, and velocity shear are obtained, reflecting the flow field stability and vortex interference intensity in the wake.

[0175] The vortex structure data obtained in the previous step, including vortex scale, curl intensity, vortex main axis direction, position center of gravity coordinates, rotation tensor principal value, etc., are used as structural dimension indicators of the state vector.

[0176] The above three types of information are constructed into a state vector according to the preset data splicing strategy. This vector serves as the input of the subsequent fuzzy neural network control model to infer the correspondence between the current state and the target control strategy.

[0177] Through this step, the conversion from physical space observation data to control space input is achieved, enabling the control model to maintain accurate response under conditions of strong disturbances and complex wake structures, and improving the operational robustness of the thruster.

[0178] S108. Input the multidimensional state vector into a preconfigured adaptive fuzzy neural network model to dynamically infer the correlation between the vortex intensity and the propeller operating state, and output the target speed adjustment amount and blade angle adjustment amount for adjusting the propeller operating state.

[0179] The preconfigured adaptive fuzzy neural network model contains multiple input membership functions, fuzzy rule sets, and output conditioning layers.

[0180] Through the fuzzy inference engine, after inputting the multi-dimensional state vector, fuzzification, rule matching and defuzzification operations are performed to obtain the control parameters for thruster adjustment.

[0181] The control parameters include:

[0182] Target speed adjustment: the speed value that instructs the propeller to increase or decrease;

[0183] Blade angle of attack adjustment: adjust the inclination angle of the propeller blades to adapt to changes in the flow field.

[0184] The system adjusts the propeller operating status in real time according to the network output to achieve the purpose of improving propulsion efficiency, reducing energy consumption and weakening adverse wake interference, thereby realizing intelligent and adaptive control.

[0185] Specifically, it includes: fuzzifying each variable in the multidimensional state vector and mapping it to a corresponding fuzzy set to form a fuzzy input variable; calculating the activation strength of each fuzzy rule based on the fuzzy input variable according to a preset fuzzy rule library; performing a comprehensive simulation of the activated fuzzy rules through the fuzzy simulation module in the adaptive fuzzy neural network model to obtain an output fuzzy set; defuzzifying the output fuzzy set to obtain a numerical target speed adjustment amount and a blade angle adjustment amount.

[0186] For each input variable (such as normalized vortex intensity, energy consumption density, propulsion efficiency, disturbance angle, etc.), a corresponding membership function set is defined; optional membership functions include triangular function, trapezoidal function, Gaussian function, etc., which are used to describe the degree of membership of the variable at different fuzzy language values ​​(such as "low", "medium", and "high");

[0187] For example, vortex intensity can be fuzzified into three linguistic values: Low, Medium, and High, corresponding to different membership function ranges. Ultimately, each input variable is mapped to one or more fuzzy input variables.

[0188] After fuzzification, the system calls the preset fuzzy rule base for reasoning. The rule base consists of several rules in the following forms:

[0189] IF vortex strength is High AND propulsion efficiency is Low AND energy density is High THEN reduce rotation speed AND increase angle of attack;

[0190] IF vortex intensity is Medium AND flow field disturbance is Low THEN maintain current speed AND fine-tune angle of attack;

[0191] For each fuzzy rule, calculate its activation strength, which is the combination of the membership degrees of all its antecedents (min or product operation can be used).

[0192] All activated fuzzy rules and their corresponding activation intensities are fed into the fuzzy simulation module within the neural network. This module has adaptive weight adjustment capabilities, dynamically adjusting the weights between rules and output offsets through the network learning process to improve inference accuracy.

[0193] By integrating the activation of multiple rules, an output fuzzy set is constructed, which forms a fuzzy description of the target output parameters (i.e., speed adjustment amount and blade angle of attack adjustment amount).

[0194] The linguistic value of the output fuzzy quantity can also be set to: Decrease, Maintain, Increase;

[0195] Each linguistic value corresponds to an output membership function, forming the final fuzzy output set.

[0196] Defuzzification is performed on the above fuzzy output set to convert the fuzzy language output into numerical control instructions. Optional defuzzification methods include:

[0197] Centroid Method: Calculates the output value based on the area centroid of the output membership function graph;

[0198] Max Membership Principle: Select the representative value corresponding to the output language value with the largest membership;

[0199] Weighted average method: The numerical mapping of multiple output language values ​​is weighted averaged according to the activation strength.

[0200] Finally, the numerical adjustment instruction is obtained:

[0201] Target speed adjustment: unit is rpm;

[0202] Blade angle of attack adjustment: the unit is angle (°).

[0203] The system transmits these two parameters to the thruster control module in real time and adjusts its current operating status to adapt to changes in vortex characteristics, suppress unstable wake interference, and optimize propulsion efficiency and energy consumption performance.

[0204] In an optional embodiment, in order to further improve the accuracy and robustness of vortex structure recognition and enhance the stability of subsequent control instruction generation, step S105 may adopt a more specific implementation method to complete the spatial aggregation and feature structure extraction of the target grid area. Figure 5 The specific implementation process of this step is described below in conjunction with an optional embodiment:

[0205] S1051. Divide multiple adjacent grid cells that meet the rotation dominance condition into a connected region as a candidate vortex region based on spatial proximity;

[0206] First, a connectivity clustering process is performed on the set of grid cells that have been selected by the rotation dominance criteria (such as the Q criterion and the λ2 criterion) in the previous stage based on their adjacency relationships in three-dimensional space. Specifically, spatially connected grid cells can be grouped into a connected subset using face adjacency or vertex adjacency criteria. Each connected subset is defined as a candidate vortex region, representing a cluster of local regions that may form vortices. In practical implementation, connected region identification algorithms such as depth-first search (DFS) or union-find can be used to implement the grid clustering process.

[0207] S1052, calculating a comprehensive vortex strength index in each candidate vortex region, wherein the comprehensive vortex strength index includes vortex core position coordinates, vortex strength, vortex core scale, and local energy consumption density;

[0208] For each candidate vortex region, further extract and calculate its internal representative physical quantities to describe its overall vortex behavior. The comprehensive vortex intensity index includes but is not limited to the following:

[0209] Vortex core position coordinates: The vortex center position is extracted through the local velocity vorticity extreme point, and the rotation axis coordinates are estimated based on the flow velocity distribution;

[0210] Vortex strength: Calculate the total rotation or integrated vorticity in the candidate region, which can be estimated using the integral form:

[0211] ;

[0212] in, is the local vorticity vector.

[0213] The vortex core scale is based on the volume of the high vorticity area within a certain threshold range from the center of the vortex core in the candidate area, and determines the radius or diameter corresponding to the vortex core;

[0214] The local energy consumption density is the volume integral of the viscous dissipation rate in the region, reflecting the degree of energy consumption of the local flow.

[0215] S1053. According to the preset vortex intensity threshold and vortex core size threshold, a target vortex region that meets the conditions of the vortex intensity threshold and the rotation dominance is screened out.

[0216] According to the set vortex intensity threshold and vortex core size threshold, each candidate vortex area is screened twice:

[0217] If the comprehensive vortex intensity of a candidate area exceeds the vortex intensity threshold and the vortex core size is within a reasonable physical range (for example, larger than the minimum numerical resolution unit and smaller than the maximum size of the tail flow), then the area is determined to be a target vortex area. At the same time, the target area must meet the verification of the conditions of rotation dominance in the early stage to ensure that its rotation characteristics are superior to other flow behaviors such as shear and diffusion.

[0218] Finally, all the physical characteristic indicators of the target vortex area are extracted to form a target multi-dimensional vortex structure characteristic data set, which serves as the input basis for subsequent control modeling and reasoning.

[0219] See Figure 6 , an embodiment of the system provided in this application is described below, and the embodiment includes:

[0220] The simulation unit 601 is used to construct a numerical simulation model of the wake region based on the geometric parameters and operating boundary conditions of the propeller, and perform simulation to obtain basic flow field data;

[0221] A tensor field unit 602 is configured to perform tensor field derivation processing based on the flow field basic data to obtain a rotation tensor and a strain tensor describing rotation and deformation characteristics;

[0222] A function mapping unit 603 is configured to map the rotation tensor and the strain tensor to a first function space and a second function space, respectively, to obtain a local tensor field index group, and to determine a target grid area based on the local tensor field index group;

[0223] a pre-screening unit 604 for generating a curl feature according to the rotation tensor, and constructing a graph structure model based on the curl feature to perform spectral clustering to obtain a candidate grid region, wherein the curl feature of the candidate grid region reaches a preset threshold;

[0224] A grid unit aggregation unit 605 is used to aggregate the grid units in the candidate grid area and determine a target vortex area, where the target vortex area has a rotation-dominant property in the candidate grid area;

[0225] The data extraction unit 606 is used to extract multi-dimensional vortex structure characteristic data including at least vortex scale, vortex intensity, rotation axis and vortex core position for the identified target vortex area;

[0226] The data acquisition unit 607 is used to collect the current propulsion efficiency index and flow field disturbance parameters, and construct a multi-dimensional state vector in combination with the multi-dimensional vortex structure characteristic data;

[0227] The model input and output unit 608 is used to input the multidimensional state vector into a preconfigured adaptive fuzzy neural network model to dynamically infer the correlation between the vortex intensity and the propeller operating state, and output the target speed adjustment amount and blade angle adjustment amount for adjusting the propeller operating state.

[0228] Optionally, the function mapping unit 603 is specifically configured to:

[0229] Mapping the rotation tensor to a first function space based on rotation characteristics, where the first function space is a characteristic mode domain constructed based on tensor eigenvalue decomposition, and is used to reflect local vortex distribution characteristics;

[0230] Mapping the strain tensor to a second function space dominated by shear deformation and principal stress directions, wherein the second function space is used to represent the anisotropic behavior of local deformation;

[0231] Extracting a first tensor response eigenvalue related to the direction of the vortex principal axis in the first function space, and extracting a second tensor response eigenvalue related to the direction of the maximum principal strain in the second function space;

[0232] constructing a local tensor field index group based on the first tensor response eigenvalue and the second tensor response eigenvalue, for characterizing the rotation dominance and shear coupling characteristics of the grid area;

[0233] A target grid region is determined based on the local tensor field indicator group.

[0234] Optionally, the function mapping unit 603 is specifically configured to:

[0235] For each grid cell in the wake region, the corresponding first tensor response eigenvalue and second tensor response eigenvalue are obtained respectively;

[0236] Based on a preset functional relationship, combining the first tensor response eigenvalue and the second tensor response eigenvalue to obtain a combined eigenvalue of each grid cell;

[0237] Screening out grid cells that do not meet the rotation dominance condition according to the relationship between the combined eigenvalue and the first threshold;

[0238] For the retained mesh elements, calculate the angle information between the main direction of the rotation tensor and the main direction of the strain tensor;

[0239] A target grid area that meets the rotation dominance condition is screened out according to the angle information.

[0240] Optionally, the tensor field unit 602 is specifically configured to:

[0241] Calculating a velocity gradient tensor for each grid cell based on a velocity vector in the flow field basic data;

[0242] Decomposing the velocity gradient tensor into an antisymmetric part and a symmetric part, wherein the antisymmetric part corresponds to a rotation tensor and is used to describe the rotation characteristics of the local fluid, and the symmetric part corresponds to a strain tensor and describes the deformation characteristics of the local fluid;

[0243] The eigenvalues ​​and eigenvectors of the rotation tensor and strain tensor are calculated respectively.

[0244] Optionally, the condition for satisfying rotation dominance includes that the combined eigenvalue is greater than or equal to the first threshold.

[0245] Optionally, the grid unit aggregation unit 606 is specifically configured to:

[0246] According to the spatial proximity relationship, multiple adjacent grid cells that meet the rotation dominance condition are divided into a connected region as a candidate vortex region;

[0247] Calculating a comprehensive vortex strength index within each candidate vortex region, wherein the comprehensive vortex strength index includes vortex core position coordinates, vortex strength, vortex core scale, and local energy consumption density;

[0248] According to the preset vortex intensity threshold and vortex core size threshold, the target vortex area that meets the conditions of vortex intensity threshold and rotation dominance is screened out.

[0249] Optionally, the model input and output unit 608 is specifically configured to:

[0250] Fuzzifying each variable in the multidimensional state vector and mapping it to a corresponding fuzzy set to form a fuzzy input variable;

[0251] According to a preset fuzzy rule base, determining the activation strength of each fuzzy rule in the fuzzy rule base based on the fuzzy input variable;

[0252] Determine a target fuzzy rule based on the activation strength, and perform comprehensive simulation on the target fuzzy rule through a fuzzy simulation module in an adaptive fuzzy neural network model to obtain an output fuzzy set;

[0253] The output fuzzy set is defuzzified to obtain a numerical target speed adjustment amount and a blade angle of attack adjustment amount.

[0254] Optionally, the pre-screening unit 604 is specifically configured to:

[0255] Calculating a curl feature of each grid cell based on the rotation tensor, wherein the curl feature is used to characterize the rotation intensity of the local flow field;

[0256] The grid cells containing the curl features are constructed into a graph structure model, in which nodes correspond to grid cells and edge weights are set based on the difference in curl features between adjacent nodes.

[0257] Applying a spectral clustering algorithm to the graph structure model and performing unsupervised partitioning on the grid cells to obtain clustering results;

[0258] Clusters whose curl characteristics exceed a preset threshold are screened out from the clustering results and determined as candidate grid areas.

[0259] See also Figure 7 , the present application also provides an underwater thruster control system based on tensor identification and fuzzy control, including:

[0260] Processor 701, memory 702, input and output unit 703, bus 704;

[0261] The processor 701 is connected to the memory 702, the input and output unit 703 and the bus 704;

[0262] The memory 702 stores a program, and the processor 701 calls the program to execute any of the above methods.

[0263] The present application also relates to a computer-readable storage medium, on which a program is stored. When the program is run on a computer, the computer is caused to execute any of the above methods.

[0264] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0265] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0266] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0267] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0268] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, read-only memory), random access memory (RAM, random access memory), disk or optical disk, and other media that can store program code.

Claims

1. An underwater thruster control method based on tensor identification and fuzzy control, characterized in that: include: Based on the propeller's geometric parameters and operating boundary conditions, a numerical simulation model of the wake region is constructed, and simulations are performed to obtain basic flow field data. Performing tensor field derivation processing based on the flow field basic data to obtain a rotation tensor and a strain tensor describing rotation and deformation characteristics; Mapping the rotation tensor and the strain tensor to a first function space and a second function space, respectively, to obtain a local tensor field index group, and determining a target grid area based on the local tensor field index group; generating a curl feature according to the rotation tensor, and constructing a graph structure model based on the curl feature to perform spectral clustering to obtain a candidate grid area, wherein the curl feature of the candidate grid area reaches a preset threshold; performing aggregation processing on each grid unit in the candidate grid area and determining a target vortex area, wherein the target vortex area has rotation dominance in the candidate grid area; For the identified target vortex region, extract multi-dimensional vortex structure characteristic data including at least vortex scale, vortex intensity, rotation axis and vortex core position; Collecting current propulsion efficiency indicators and flow field disturbance parameters, and constructing a multidimensional state vector in combination with the multidimensional vortex structure characteristic data; The multidimensional state vector is input into a preconfigured adaptive fuzzy neural network model to dynamically infer the correlation between the vortex intensity and the propeller operating state, and the target speed adjustment amount and blade angle adjustment amount for adjusting the propeller operating state are output.

2. The underwater thruster control method based on tensor identification and fuzzy control according to claim 1, characterized in that: Generating a curl feature according to the rotation tensor, and constructing a graph structure model based on the curl feature to perform spectral clustering to obtain a candidate grid area, wherein the curl feature of the candidate grid area reaches a preset threshold, including: Calculating a curl feature of each grid cell based on the rotation tensor, wherein the curl feature is used to characterize the rotation intensity of the local flow field; The grid cells containing the curl features are constructed into a graph structure model, in which nodes correspond to grid cells and edge weights are set based on the difference in curl features between adjacent nodes. Applying a spectral clustering algorithm to the graph structure model and performing unsupervised partitioning on the grid cells to obtain clustering results; Clusters whose curl characteristics exceed a preset threshold are screened out from the clustering results and determined as candidate grid areas.

3. The underwater thruster control method based on tensor identification and fuzzy control according to claim 1, characterized in that: Mapping the rotation tensor and the strain tensor to the first function space and the second function space respectively to obtain a local tensor field index group, and determining the target grid area based on the local tensor field index group includes: Mapping the rotation tensor to a first function space based on rotation characteristics, where the first function space is a characteristic mode domain constructed based on tensor eigenvalue decomposition, and is used to reflect local vortex distribution characteristics; Mapping the strain tensor to a second function space dominated by shear deformation and principal stress directions, wherein the second function space is used to represent the anisotropic behavior of local deformation; Extracting a first tensor response eigenvalue related to the direction of the vortex principal axis in the first function space, and extracting a second tensor response eigenvalue related to the direction of the maximum principal strain in the second function space; constructing a local tensor field index group based on the first tensor response eigenvalue and the second tensor response eigenvalue, for characterizing the rotation dominance and shear coupling characteristics of the grid area; A target grid region is determined based on the local tensor field indicator group.

4. The underwater thruster control method based on tensor identification and fuzzy control according to claim 3 is characterized in that: Determining the target grid area based on the local tensor field indicator group includes: For each grid cell in the wake region, the corresponding first tensor response eigenvalue and second tensor response eigenvalue are obtained respectively; Based on a preset functional relationship, combining the first tensor response eigenvalue and the second tensor response eigenvalue to obtain a combined eigenvalue of each grid cell; Screening out grid cells that do not meet the rotation dominance condition according to the relationship between the combined eigenvalue and the first threshold; For the retained mesh elements, calculate the angle information between the main direction of the rotation tensor and the main direction of the strain tensor; A target grid area that meets the rotation dominance condition is screened out according to the angle information.

5. The underwater thruster control method based on tensor identification and fuzzy control according to claim 1, characterized in that: The tensor field derivation process is performed based on the flow field basic data to obtain the rotation tensor and strain tensor describing the rotation and deformation characteristics, including: Calculating a velocity gradient tensor for each grid cell based on a velocity vector in the flow field basic data; Decomposing the velocity gradient tensor into an antisymmetric part and a symmetric part, wherein the antisymmetric part corresponds to a rotation tensor and is used to describe the rotation characteristics of the local fluid, and the symmetric part corresponds to a strain tensor and describes the deformation characteristics of the local fluid; The eigenvalues ​​and eigenvectors of the rotation tensor and strain tensor are calculated respectively.

6. The underwater thruster control method based on tensor identification and fuzzy control according to claim 4 is characterized in that: The condition for satisfying the rotation dominance includes that the combined eigenvalue is greater than or equal to the first threshold.

7. The underwater thruster control method based on tensor identification and fuzzy control according to claim 1, characterized in that: Aggregating each grid cell in the candidate grid area and determining a target vortex area, wherein the target vortex area has rotation dominance in the candidate grid area, comprises: According to the spatial proximity relationship, multiple adjacent grid cells that meet the rotation dominance condition are divided into a connected region as a candidate vortex region; Calculating a comprehensive vortex strength index within each candidate vortex region, wherein the comprehensive vortex strength index includes vortex core position coordinates, vortex strength, vortex core scale, and local energy consumption density; According to the preset vortex intensity threshold and vortex core size threshold, the target vortex area that meets the conditions of vortex intensity threshold and rotation dominance is screened out.

8. The underwater thruster control method based on tensor identification and fuzzy control according to claim 1, characterized in that: Inputting the multidimensional state vector into a preconfigured adaptive fuzzy neural network model to dynamically infer the correlation between the vortex intensity and the propeller operating state, and outputting a target speed adjustment amount and a blade angle of attack adjustment amount for adjusting the propeller operating state includes: Fuzzifying each variable in the multidimensional state vector and mapping it to a corresponding fuzzy set to form a fuzzy input variable; According to a preset fuzzy rule base, determining the activation strength of each fuzzy rule in the fuzzy rule base based on the fuzzy input variable; Determine a target fuzzy rule based on the activation strength, and perform comprehensive simulation on the target fuzzy rule through a fuzzy simulation module in an adaptive fuzzy neural network model to obtain an output fuzzy set; The output fuzzy set is defuzzified to obtain a numerical target speed adjustment amount and a blade angle of attack adjustment amount.

9. Underwater thruster control system based on tensor identification and fuzzy control, characterized in that: include: A simulation unit is used to construct a numerical simulation model of the wake region based on the geometric parameters and operating boundary conditions of the propeller, and to perform simulation to obtain basic flow field data; A tensor field unit is used to perform tensor field derivation processing based on the flow field basic data to obtain a rotation tensor and a strain tensor describing rotation and deformation characteristics; a function mapping unit, configured to map the rotation tensor and the strain tensor to a first function space and a second function space, respectively, to obtain a local tensor field index group, and to determine a target grid area based on the local tensor field index group; a pre-screening unit, configured to generate a curl feature according to the rotation tensor, and construct a graph structure model based on the curl feature to perform spectral clustering to obtain a candidate grid area, wherein the curl feature of the candidate grid area reaches a preset threshold; a grid unit aggregation unit for performing aggregation processing on each grid unit in the candidate grid area and determining a target vortex area, wherein the target vortex area has rotation dominance in the candidate grid area; A data extraction unit is used to extract multi-dimensional vortex structure characteristic data including at least vortex scale, vortex intensity, rotation axis and vortex core position from the identified target vortex area; A data acquisition unit, configured to collect current propulsion efficiency indicators and flow field disturbance parameters, and construct a multidimensional state vector based on the multidimensional vortex structure characteristic data; The model input and output unit is used to input the multidimensional state vector into a preconfigured adaptive fuzzy neural network model to dynamically infer the correlation between the vortex intensity and the propeller operating state, and output the target speed adjustment amount and blade angle of attack adjustment amount for adjusting the propeller operating state.

10. Underwater propulsion control system based on tensor identification and fuzzy control, characterized in that: include: processor, memory, input and output units, and buses; The processor is connected to the memory, the input and output unit, and the bus; The memory stores a program, and the processor calls the program to execute the method according to any one of claims 1 to 8.

Images