Intelligent flow field shock wave identification and visualization method

Abstract

The invention relates to the technical field of flow field shock wave recognition and visualization methods, in particular to an intelligent flow field shock wave recognition and visualization method, which comprises a data preprocessing module used for preprocessing original flow field data to obtain normalized input data in a multidimensional tensor form; the feature extraction module is used for extracting local gradient features of input data through special matrix mapping so as to generate a preliminary feature matrix; the topological association module constructs a topological structure based on the preliminary feature matrix and generates a topological feature matrix; the attention enhancement module is used for weighting the topological feature matrix through a special function so as to enhance the significant features of the shock wave region; the output module generates a final shock wave characteristic matrix based on the weighted matrix for visual presentation, a special matrix method is introduced for characteristic extraction, and change details of a shock wave area in a flow field can be accurately captured by combining periodicity and gradient characteristics of a trigonometric function.

Description

Technical Field

[0001] The invention relates to the technical field of flow field shock wave identification and visualization methods, and in particular to an intelligent flow field shock wave identification and visualization method. Background Art

[0002] With the rapid development of fluid mechanics and computational fluid dynamics (CFD) technology, flow field analysis and shock wave identification have been widely used in high-tech fields such as aerospace, automotive industry and turbomachinery. Accurate identification of shock waves in the flow field is of great significance for optimizing design, improving system efficiency, reducing flow losses and ensuring operational safety. However, due to the complex characteristics of flow field shock waves, existing technologies face many challenges in identifying and processing shock wave characteristics.

[0003] At present, shock wave identification mainly relies on numerical methods or manual intervention. For example, traditional numerical methods identify the shock wave position by setting a threshold to judge the pressure gradient or density gradient in the flow field. These methods have a certain theoretical basis, but there are obvious limitations in their practical applications. First, the characteristics of flow field shock waves show great differences under different flow conditions, and a single threshold method is difficult to fully adapt to the dynamic changes of complex flow fields; second, the threshold method often relies on human experience settings, which is highly subjective and prone to misjudgment or missed judgments, especially in complex flow fields with multiple shock waves and multiple structures. The misjudgment rate increases significantly; third, single feature extraction based on gradient changes is difficult to capture the global characteristics of the shock wave area, and lacks in-depth analysis of the integrity of the flow field and the correlation between shock waves.

[0004] At the same time, with the improvement of flow field simulation accuracy, the scale of flow field data has grown exponentially, and traditional manual intervention methods have gradually exposed problems of low efficiency and poor adaptability. In the existing technology, some methods have tried to introduce automated algorithms, such as image processing or simple machine learning models, but these methods are mostly limited to feature point extraction and fail to fully enhance the significant characteristics of the shock wave area. In addition, existing automated methods are usually difficult to process large-scale flow field data, the recognition process lacks real-time performance, and does not match the dynamic characteristics of complex flow fields.

[0005] Existing technologies also have shortcomings in the field of visualization. Shock wave identification and visualization are often separated into independent steps, requiring multiple data transfers to complete identification and display. This model is prone to data loss or error accumulation, resulting in insufficient accuracy and clarity of the final visualization results. In addition, due to the independence of identification and visualization methods, the efficiency of system integration and ease of operation are also limited.

[0006] Specifically, the present invention uses a data preprocessing module to standardize and convert the raw flow field data into multi-dimensional tensors to ensure the high quality and consistency of data input; the feature extraction module uses a special matrix to capture and map local features of normalized data, significantly improving the sensitivity to shock wave area features; the topological association module further globally associates local features by constructing a topological structure, solving the problem of lack of coupling between local features and global information in traditional methods.

[0007] The technical problem solved by the present invention is how to achieve efficient and accurate recognition of shock wave characteristics under complex flow field conditions and display the recognition results in a clear and intuitive manner, thereby providing more efficient and reliable technical support for fluid mechanics analysis and engineering optimization. Summary of the invention

[0008] Based on the defects and deficiencies of the above-mentioned prior art, the present invention proposes an intelligent flow field shock wave recognition and visualization method to solve the technical problems existing in the existing methods. Through the hierarchical design and synergy of the data preprocessing module, feature extraction module and topological association module, the present invention realizes the automation of the whole process from flow field data preprocessing to shock wave feature extraction and global topological association. At the same time, combined with innovative algorithms, it overcomes the problems of low recognition accuracy, poor adaptability and lack of global information in the existing methods.

[0009] The present invention provides an intelligent flow field shock wave identification and visualization method, comprising:

[0010] A data preprocessing module is used to preprocess the original flow field data to obtain normalized input data in the form of multi-dimensional tensors;

[0011] A feature extraction module extracts local gradient characteristics of the input data through special matrix mapping to generate a preliminary feature matrix;

[0012] A topological association module, which constructs a topological structure based on the preliminary feature matrix and generates a topological feature matrix; an attention enhancement module, which weights the topological feature matrix through a special function to enhance the significant features of the shock wave area;

[0013] The output module generates a final shock wave characteristic matrix based on the weighted matrix for visual presentation. Preferably, the data preprocessing module includes:

[0014] The original data X is calculated based on the following formula raw Perform normalization:

[0015]

[0016] Among them, X norm is the normalized data matrix, min(Xraw ) and max(X raw ) are the minimum and maximum values ​​of the original data respectively;

[0017] Perform a multidimensional tensor transformation on the normalized data to obtain data in a multidimensional tensor form:

[0018] T norm = reshape(X norm , d1, d2, d3)

[0019] Among them, T norm It is a multidimensional tensor, d1, d2, d3 are the spatial and feature dimensions of the tensor.

[0020] Preferably, the feature extraction module comprises:

[0021] The following formula is used to extract features from multidimensional tensor data:

[0022] F=M·T norm

[0023] Among them, F is the characteristic matrix and M is the special matrix, which are defined as follows:

[0024]

[0025] T norm It is the output tensor of the data preprocessing module.

[0026] Preferably, the topology association module includes:

[0027] A topological feature matrix is ​​generated for the feature matrix F based on the following formula:

[0028] T topo =f topo (F)

[0029]

[0030] Among them, T topo is the topological characteristic matrix, is the neighborhood set of node i, ||F k，j ||2 is F k，j Preferably, the attention enhancement module comprises:

[0031] A i，j =φ((T topo，i，j )·T topo,i，j

[0032]

[0033] Among them, A i，jis the matrix element after attention enhancement, φ(T topo，i，j ) is a special function used to calculate the attention weight.

[0034] Preferably, the output module comprises:

[0035] The final shock wave characteristic matrix is ​​generated based on the following formula:

[0036] O=A·W

[0037]

[0038] Among them, O is the shock wave characteristic matrix, W is the dynamically updated weight matrix, ||A i ,:||1 is the L1 norm of the i-th row of matrix A.

[0039] Preferably, the special function combines an exponential function and a Sigmoid function to enhance the significant features of the shock wave region.

[0040] Preferably, the output weight matrix is ​​dynamically updated by a back-propagation algorithm to achieve adaptive optimization of shock wave characteristics.

[0041] Preferably, the feature extraction module and the attention enhancement module have a synergistic effect, and the output of the feature matrix is ​​optimized through multiple iterations to finally generate a shock wave feature matrix.

[0042] Preferably, the method is suitable for real-time identification and dynamic visualization of shock waves in complex flow fields.

[0043] The beneficial effects of the present invention are:

[0044] The method of the present invention is designed in layers through a data preprocessing module, a feature extraction module and a topological association module, so that the output of each step can be used as an efficient input for the next step, ensuring the logic and consistency of the entire method. The normalized tensor form data provided by the data preprocessing module not only lays the foundation for the feature extraction of the subsequent modules, but also effectively solves the problem of inconsistent scales of flow field data; the feature extraction module significantly improves the expression ability of local features through the mapping operation of a special matrix, and the topological association module further expands the local features into global features through a global topological structure, thereby achieving complementarity, superposition and coordination between steps.

[0045] The present invention introduces a special matrix method for feature extraction. By combining the periodicity and gradient characteristics of trigonometric functions, it can accurately capture the details of changes in the shock wave area in the flow field; the topological association module ensures the integrity and consistency of the global features based on the accumulation and normalization of neighborhood features. The entire method not only improves the recognition accuracy but also enhances the adaptability of the algorithm in complex flow fields through input-output association and progressive optimization between algorithms.

[0046] The method of the present invention realizes an integrated technical process from flow field data preprocessing to shock wave feature extraction and then to global association, avoiding the data transmission error and inefficiency caused by step separation in traditional methods. In addition, through the global information enhancement of the topological association module, the method of the present invention significantly improves the global accuracy and visualization effect of shock wave identification.

[0047] Preferably, the method of the present invention is applicable to shock wave identification in various complex flow field environments, and can achieve efficient identification of weak shock waves in low-speed flow fields and strong shock wave features in ultra-high-speed flow fields. This applicability is due to the normalization of flow field data by the data preprocessing module, and the highly adaptable algorithm design of the feature extraction and topological association modules.

[0048] In summary, the present invention proposes an intelligent flow field shock wave identification and visualization method. Through innovative modular design and algorithm optimization, it not only solves the problems of low recognition accuracy, poor adaptability, and insufficient global features in the prior art, but also significantly improves the efficiency and accuracy of flow field shock wave identification and visualization, providing technical support and practical guidance for related fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a logic block diagram of the method of the present invention.

[0050] Figure 2 It is a logic block diagram of the data preprocessing module of the present invention.

[0051] Figure 3 It is a logic block diagram of the feature extraction module of the present invention.

[0052] Figure 4 It is a logic block diagram of the topology association module of the present invention.

[0053] Figure 5 It is a flow chart of the output logic of the present invention. DETAILED DESCRIPTION

[0054] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects thereof are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0055] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0056] The technical solution of the present invention is described in detail below in conjunction with specific embodiments.

[0057] Please refer to Figure 1-5 The present invention provides an intelligent flow field shock wave identification and visualization method, including a data preprocessing module 1, a feature extraction module 2 and a topological association module 3. Through the synergistic effect of the above modules, the method can achieve efficient identification and accurate visualization of shock wave features in complex flow fields. Specifically, the data preprocessing module 1 normalizes the original flow field data and converts it into a multi-dimensional tensor form to ensure the consistency and standardization of the data, providing high-quality input for the calculation of subsequent modules. The feature extraction module 2 uses a special matrix to linearly map the normalized data to extract local gradient features and generate a preliminary feature matrix. Subsequently, the topological association module 3 further associates the local information of the feature matrix to the global space by constructing a topological structure, thereby generating a topological feature matrix.

[0058] Preferably, in one embodiment of the present invention, in the data preprocessing module 1, the normalization operation of the original flow field data adopts the following formula:

[0059]

[0060] Among them, X raw is the original flow field data matrix, representing the input data obtained by sampling; X norm is the normalized data matrix, indicating the standardized data; min(X raw ) and max(X raw ) are the minimum and maximum values ​​of the data matrix, respectively.

[0061] Furthermore, in order to meet the requirements of representing multi-dimensional flow field characteristics, this method performs multi-dimensional tensor transformation on the normalized data. The specific formula is as follows:

[0062] T norm = reshape(X norm , d1, d2, d3)

[0063] Among them, T norm It is normalized data in the form of a multi-dimensional tensor, where d1, d2, and d3 are three dimensions of the tensor. Preferably, d1 and d2 are spatial dimensions, and d3 is a feature dimension.

[0064] This step not only significantly improves the consistency of the data through standardization and tensorization, but also lays a solid foundation for feature extraction in subsequent modules. The normalization range [0,1] ensures that the scale differences between different physical quantities are eliminated, thereby avoiding numerical deviation problems in subsequent processing.

[0065] In one embodiment of the present invention, the feature extraction module 2 extracts features from the normalized multidimensional tensor data through special matrix mapping to generate a preliminary feature matrix. The core innovation of this module is to use a special matrix to effectively capture the local gradient characteristics of the flow field data and generate a preliminary feature matrix that can characterize the local characteristics of the flow field.

[0066] Preferably, in one embodiment of the present invention, the specific formula of the feature extraction process is as follows:

[0067] F=M·T norm

[0068] Among them, F is the preliminary feature matrix, and M is the special matrix of feature mapping, which is defined as:

[0069]

[0070] T norm It is the multi-dimensional tensor output by the data preprocessing module 1.

[0071] The special matrix M in the above formula is one of the core innovations of the present invention. By combining the periodicity and gradient variation characteristics of trigonometric functions, it can effectively capture the local changes in the shock wave area in the flow field data.

[0072] Preferably, in one embodiment of the present invention, the dimension of the matrix M is dynamically generated according to the size of the input tensor to ensure that the information in each tensor channel can be fully mapped. The design of the feature extraction module 2 not only simplifies the complexity of data processing, but also significantly improves the sensitivity to the local characteristics of the flow field through the linear transformation of the special matrix. For example, when the spatial dimension of the input data is 100×100, the size of the matrix M is 100×100, and its computational complexity is optimized to O(n 2 ), which greatly improves the computational efficiency.

[0073] In one embodiment of the present invention, the topological association module 3 further enhances the feature matrix by constructing a topological structure to generate a topological feature matrix. The main function of this module is to associate the local information of the feature matrix with the global topological structure in order to capture the global shock wave characteristics in the flow field.

[0074] Preferably, in one embodiment of the present invention, the formula for generating the topological feature is as follows:

[0075] T topo =f topo (F)

[0076]

[0077] Among them, T topo is the topological characteristic matrix, f topo is the feature enhancement function; F k，j is the element in the i-th row and j-th column of the feature matrix F; is the neighborhood set of node i, representing all nodes connected to node i; ||F k，j ||2 is F k，j The L2 norm of is used for normalization. This module realizes global enhancement of features through the cumulative association of neighborhood information, and is particularly suitable for processing shock wave regions with complex structures and strong nonlinear changes in the flow field. Preferably, when the neighborhood size When the number of nodes is set to 8 (i.e., eight-neighborhood mode), the sensitivity of the topological feature matrix to the shock wave region can be significantly improved.

[0078] The beneficial effect of this step is that a characteristic matrix that can reflect the global shock wave structure is constructed through topological association, which effectively overcomes the error accumulation problem easily caused by traditional methods that only rely on local features, and provides rich input information for the recognition and visualization of subsequent modules.

[0079] The method of the present invention realizes the complete conversion process from raw flow field data to topological feature matrix through layer-by-layer processing of data preprocessing module 1, feature extraction module 2 and topological association module 3. Each module is closely related, and the output is used as the input of the next module, ensuring the consistency and hierarchy of the algorithm logic. Through normalization, multidimensional tensor transformation, special matrix mapping and topological association, the method makes full use of the physical characteristics and mathematical structure of flow field data, and provides an efficient and accurate solution for intelligent flow field shock wave recognition and visualization.

[0080] In one embodiment of the present invention, the attention enhancement module of the present invention weights the topological feature matrix through a special function to enhance the significant features of the shock wave region and generate a weighted attention matrix. The design of this module is intended to highlight the shock wave characteristic region from the complex flow field topology structure, thereby further improving the accuracy and efficiency of subsequent shock wave identification and visualization.

[0081] Preferably, in one embodiment of the present invention, the weighted operation in the attention enhancement module is implemented by the following formula:

[0082] A i，j =φ(T topo，i，j )·T topo，j

[0083] Among them, A i，j is the element in the weighted attention matrix; T topo，i，j is the topological feature matrix T topo The element in the i-th row and j-th column of topo，i，j ) is a special function for calculating attention weights, defined as follows:

[0084]

[0085] The above special function combines the characteristics of the exponential function and the square exponential function, which can effectively amplify the high-intensity features in the shock wave area and suppress the noise features in the non-shock wave area. The above special function combines the characteristics of the exponential function and the square exponential function, which can effectively amplify the high-intensity features in the shock wave area and suppress the noise features in the non-shock wave area.

[0086] Preferably, in one embodiment of the present invention, in order to ensure the consistency of the results after attention weighting, the weighted operation adopts an element-by-element processing method, and the weight calculation processes the elements of each topological feature independently to avoid the numerical overflow problem in high-dimensional matrix calculation. Furthermore, when the dimension of the topological feature matrix is ​​large, the attention enhancement module can perform block operations on the matrix through a block processing mechanism, thereby reducing the computational complexity and improving the computational efficiency.

[0087] Through the above method, the attention enhancement module effectively highlights the shock wave characteristic area and suppresses the feature redundancy of irrelevant areas, thus laying a solid foundation for the subsequent output of shock wave features.

[0088] In one embodiment of the present invention, the output module of the present invention generates the final shock wave feature matrix through a dynamic weight matrix. The design of this module combines the characteristics of the attention enhancement matrix and introduces a dynamic weight update mechanism to ensure that the generated shock wave feature matrix can adapt to the shock wave characteristics of different flow fields.

[0089] Preferably, in one embodiment of the present invention, the generation formula of the final shock wave characteristic matrix is ​​as follows:

[0090] O=A·W

[0091] Among them, O is the final shock wave feature matrix; A is the attention enhancement matrix, which is output by the attention enhancement module; W is the dynamic weight matrix, which is used to adjust the weight distribution of shock wave features. The specific calculation formula of the weight matrix W is as follows:

[0092]

[0093] Among them, W i，j is the element in the weight matrix; A i , j is an element in the attention enhancement matrix; ||A i，： ||1 is the L1 norm of the i-th row of matrix A.

[0094] Preferably, in order to ensure the calculation stability of the shock wave characteristic matrix, in one embodiment of the present invention, a logarithmic function and L1 norm normalization operation are introduced into the calculation of the weight matrix, thereby avoiding the calculation instability problem caused by excessively large values. At the same time, by dynamically adjusting the distribution of the weight matrix, the final shock wave characteristic matrix can be made to better fit the characteristics of the shock wave region in a complex flow field environment.

[0095] Through the above design, the output module can adaptively adjust the significance weights of shock wave features, thereby ensuring that the generated shock wave feature matrix has high robustness and adaptability, providing high-quality data input for subsequent visualization presentation.

[0096] In one embodiment of the present invention, the method of the present invention further optimizes the feature expression of the shock wave region through the synergy between the feature extraction module and the attention enhancement module. The synergy design combines the preliminary feature matrix of the feature extraction module and the significance weighting characteristics of the attention enhancement module, thereby significantly improving the accuracy of shock wave recognition.

[0097] Preferably, in one embodiment of the present invention, the preliminary feature matrix F generated in the feature extraction module and the attention matrix A generated in the attention enhancement module are collaboratively optimized by the following formula:

[0098] F′=A·F

[0099] Among them, F′ is the feature matrix after collaborative optimization; F is the preliminary feature matrix of the feature extraction module; A is the attention enhancement matrix. The collaboratively optimized feature matrix F′ has both the local characteristics of the feature extraction module and the global saliency characteristics of the attention enhancement module, thus providing more accurate input for the calculation of the topological association module.

[0100] Furthermore, the method of the present invention preferably iterates the synergy of the feature matrix and the attention matrix multiple times through an iterative optimization mechanism until the change amplitude of the feature matrix meets the convergence condition. The specific convergence condition is:

[0101] ||F′-F||2<∈

[0102] Where ||F′-F||2 is the L2 norm difference before and after the feature matrix is ​​updated; ∈ is the convergence threshold, preferably, ∈ is set to 10 -5 .

[0103] Through the coordinated optimization of the feature extraction module and the attention enhancement module, the method of the present invention can further improve the accuracy and stability of the feature matrix, providing stronger technical support for the accurate identification of shock wave areas in complex flow fields.

[0104] In one embodiment of the present invention, the method of the present invention further optimizes the synergy between the topological association module and the attention enhancement module, so that the generation process of the shock wave feature matrix has higher global consistency and regional adaptability.

[0105] Preferably, in one embodiment of the present invention, the topological feature matrix T generated by the topological association module topo The attention matrix A generated by the attention enhancement module is fused through the following formula:

[0106] T′=A·T topo

[0107] Among them, T′ is the fused topological feature matrix; A is the attention enhancement matrix; T topo It is the topological feature matrix output by the topological association module.

[0108] Furthermore, in order to ensure the global consistency of the topological feature matrix, a regularization operation is introduced into the fused feature matrix T′ in one embodiment of the present invention, and the regularization formula is:

[0109]

[0110] Where T″ is the regularized topological feature matrix; ||T′||1 is the L1 norm of the matrix T′.

[0111] The above method performs saliency weighting and regularization on the topological feature matrix through the attention enhancement module, so that the generated feature matrix can not only accurately describe the salient characteristics of the shock wave region, but also effectively suppress noise interference in the global range, thereby further improving the expressiveness and robustness of the shock wave feature matrix.

[0112] The method of the present invention realizes efficient recognition of shock wave regions in complex flow fields through the synergy between the modules. Through attention enhancement, dynamic weight update and module collaborative optimization, the method can not only capture the significant characteristics of the shock wave region, but also achieve the consistency of global features and the adaptability of regional features in complex environments, significantly improving the accuracy and efficiency of shock wave recognition.

[0113] In one embodiment of the present invention, the method of the present invention further optimizes the generation process of the shock wave characteristic matrix through a dynamically updated weight matrix, so that it can adapt to the complex shock wave characteristics of different flow field conditions. The update mechanism of the weight matrix is ​​based on the back propagation algorithm, and the significance of the shock wave characteristics and the consistency of the global characteristics are ensured by dynamically adjusting the distribution of weights.

[0114] Preferably, in one embodiment of the present invention, the dynamic update of the weight matrix W is implemented using the following formula:

[0115]

[0116] Among them, W (t) is the weight matrix at the tth iteration; η is the learning rate, which is used to control the step size of weight update. Preferably, the learning rate is set to 0.001; L(W (t) ) is the loss function, which represents the optimization target of the current weight matrix. Its specific form is as follows:

[0117]

[0118] Among them, O i，j is the element in the shock wave characteristic matrix, T target，i，j is an element in the target feature matrix, which is used to guide the weight update. Preferably, in one embodiment of the present invention, after each weight matrix update, the matrix is ​​normalized to ensure the numerical stability and computational efficiency of the weights. The normalization formula is as follows:

[0119]

[0120] Among them, W′ is the normalized weight matrix, and ||W||1 is the L1 norm of the weight matrix.

[0121] Through the dynamic update mechanism, the weight matrix can adjust its distribution in real time according to the changes in flow field conditions, thereby ensuring that the generated shock wave feature matrix has high robustness and adaptability in different complex environments. This adaptive optimization mechanism effectively solves the problem that the traditional fixed weight method performs poorly in complex flow fields, and provides a more flexible and efficient solution for shock wave identification.

[0122] In one embodiment of the present invention, the method of the present invention further optimizes the generation process of the shock wave feature matrix through the synergistic effect of the feature extraction module and the attention enhancement module, so that the feature matrix can simultaneously have the ability to describe the details of local features and the ability to express the significance of global features.

[0123] Preferably, in one embodiment of the present invention, the preliminary feature matrix F generated by the feature extraction module and the attention matrix A generated by the attention enhancement module are collaboratively optimized by the following formula:

[0124] F′=A·F

[0125] Among them, F′ is the optimized feature matrix; F is the preliminary feature matrix, which comes from the feature extraction module; A is the attention matrix, which comes from the attention enhancement module.

[0126] Furthermore, in order to ensure that the feature matrix after collaborative optimization has global consistency, the method of the present invention uses the following regularization formula to normalize the optimized feature matrix:

[0127]

[0128] Wherein, F″ is the regularized feature matrix; ||F′||1 is the L1 norm of the matrix F′. Preferably, in one embodiment of the present invention, in order to improve the efficiency and accuracy of collaborative optimization, a multiple iterative update mechanism is introduced in the optimization process. The termination condition of the iterative update is that the L2 norm of the matrix change is less than the set convergence threshold ∈, that is:

[0129] ||F (t+1) -F (t) ||2<∈

[0130] Where, ∈ is the convergence threshold, preferably set to 10 -6 .

[0131] Through the synergy of the feature extraction module and the attention enhancement module, this method can significantly improve the global description ability and local detail capture ability of the feature matrix, providing a more accurate and robust solution for shock wave identification in complex flow field environments.

[0132] In one embodiment of the present invention, the method of the present invention further optimizes the entire process of generating and dynamically visualizing the shock wave feature matrix, ensuring real-time identification and intuitive display of the shock wave region in a complex flow field environment. The method of the present invention achieves an efficient closed loop from flow field data acquisition to visualization result output through dynamic updating, collaborative optimization, and visualization rendering of the feature matrix.

[0133] Preferably, in one embodiment of the present invention, the finally generated shock wave feature matrix O is connected to the input of the visualization rendering module through the following formula:

[0134] V=render(O,α)

[0135] Among them, V is the visualization result after rendering; O is the final shock wave characteristic matrix; α is a transparency parameter used to control the visualization effect of the shock wave area. Preferably, the transparency parameter is set to 0.8.

[0136] Furthermore, in order to improve the efficiency and effect of visualization rendering, the method of the present invention adopts a progressive update mechanism in the rendering process. Specifically, the visualization result is updated by rendering frame by frame, and the calculation formula for each frame is as follows:

[0137] V (t+1) =β·V (t) +(1-β)·render(O (t+1) ,α)

[0138] Among them, V (t) is the visualization result of the tth frame; β is the update weight, preferably, it is set to 0.7.

[0139] Through the above dynamic visualization mechanism, the method of the present invention can update the visualization results of the shock wave area in real time when the flow field data changes, ensuring that the user can clearly and intuitively observe the dynamic characteristics of the shock wave. This real-time linkage mechanism of rendering and data calculation significantly improves the overall efficiency of shock wave identification and visualization.

[0140] The method of the present invention constructs a complete technical process from flow field data acquisition to shock wave feature recognition and then to visualization result display through dynamic weight update, module collaborative optimization and dynamic visualization rendering. This method can not only adapt to the changes in shock wave characteristics under complex flow field conditions, but also provide real-time visualization results through an efficient dynamic rendering mechanism, significantly improving the overall performance of shock wave recognition and visualization, and has extremely high practical value and innovation.

[0141] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent flow field shock wave identification and visualization method, characterized in that: include: A data preprocessing module is used to preprocess the original flow field data to obtain normalized input data in the form of multi-dimensional tensors; A feature extraction module extracts local gradient characteristics of the input data through special matrix mapping to generate a preliminary feature matrix; A topological association module, which constructs a topological structure based on the preliminary feature matrix and generates a topological feature matrix; An attention enhancement module, which weights the topological feature matrix through a special function to enhance the salient features of the shock wave region; The output module generates the final shock wave characteristic matrix based on the weighted matrix for visualization.

2. The intelligent flow field shock wave identification and visualization method according to claim 1, characterized in that: The data preprocessing module comprises: The original data X is calculated based on the following formula raw Perform normalization: Among them, X norm is the normalized data matrix, min(X raw ) and max(X raw ) are the minimum and maximum values ​​of the original data respectively; Perform a multidimensional tensor transformation on the normalized data to obtain data in a multidimensional tensor form: T norm =reshape(X norm ,d1,d2,d3) Among them, T norm It is a multidimensional tensor, d1, d2, d3 are the spatial and feature dimensions of the tensor.

3. The intelligent flow field shock wave identification and visualization method according to claim 1, characterized in that: The feature extraction module comprises: The following formula is used to extract features from multidimensional tensor data: F=M·T norm Among them, F is the characteristic matrix and M is the special matrix, which are defined as follows: T norm It is the output tensor of the data preprocessing module.

4. The intelligent flow field shock wave identification and visualization method according to claim 1, characterized in that: The topology association module includes: A topological feature matrix is ​​generated for the feature matrix F based on the following formula: T topo =f topo (F) Among them, T topo is the topological characteristic matrix, is the neighborhood set of node i, ||F k，j ||2 is F k，j The L2 norm of .

5. The intelligent flow field shock wave identification and visualization method according to claim 1, characterized in that: The attention enhancement module comprises: A i,j =φ(T topo，i，j )·T topo，i，j Among them, A i，j is the matrix element after attention enhancement, φ(T topo，i，j ) is a special function used to calculate the attention weight.

6. The intelligent flow field shock wave identification and visualization method according to claim 1, characterized in that: The output module comprises: The final shock wave characteristic matrix is ​​generated based on the following formula: O=A·W Among them, O is the shock wave characteristic matrix, W is the dynamically updated weight matrix, ||A i ,:|| I is the L of the i-th row of matrix A I Norm.

7. The intelligent flow field shock wave identification and visualization method according to claim 5, characterized in that: The special function combines the exponential function and the Sigmoid function to enhance the significant features of the shock wave region.

8. The intelligent flow field shock wave identification and visualization method according to claim 6, characterized in that: The output weight matrix is ​​dynamically updated through a back-propagation algorithm to achieve adaptive optimization of shock wave characteristics.

9. The intelligent flow field shock wave identification and visualization method according to claim 1, characterized in that: The feature extraction module and the attention enhancement module have a synergistic effect, and the output of the feature matrix is ​​optimized through multiple iterations to finally generate a shock wave feature matrix.

10. The intelligent flow field shock wave identification and visualization method according to claim 1, characterized in that: The method is suitable for real-time identification and dynamic visualization of shock waves in complex flow fields.

Images