Visualization Acceleration Method and System for Particle-like Algorithms Based on Clustering

Abstract

The present invention relates to the field of explosion damage simulation, and particularly to a visualization acceleration method and system for a particle-based algorithm based on clustering. The method includes: obtaining a data result file of explosion damage; obtaining the sampling scene features of the data result file, and finding a downsampling method that matches the sampling scene features in the clustering downsampling method library according to a preset matching rule; selecting a first clustering condition that matches the sampling scene features and the clustering downsampling method from the first clustering condition database through the sampling scene features and the downsampling method; selecting a second clustering condition input by the user through the downsampling method; and performing clustering downsampling on the visualization physical features of the data result file by using the first and second clustering conditions through the selected downsampling method. For application scenarios such as explosion damage simulation, the present invention can analyze large-volume data result files within a limited time to meet the business requirements of users such as real-time monitoring, interactive analysis, and decision support.

Description

Technical Field

[0001] The present invention relates to the field of explosion damage simulation, and particularly to a visualization acceleration method and system for a particle-based algorithm based on clustering. Background Art

[0002] Visual simulation is one of the key means for studying the fracture, explosion and impact problems of materials. At present,

[0003] The visual simulation of explosion damage is mainly divided into two types of routes:

[0004] 1) Traditional grid simulation calculation

[0005] Grid simulation calculation usually uses grid data as the calculation object to numerically simulate the fracture and explosion damage of materials, and then presents the fracture and explosion damage of materials in a visual form.

[0006] For example, the invention patent application with the patent application number CN202011250264.4 discloses a visualization simulation analysis method for the explosion of dangerous goods in a large-space complex building structure. This method establishes a three-dimensional model of the large-space complex building structure and its surrounding environment; then analyzes the damage and overpressure conditions of the dangerous source room and its surrounding rooms, and represents the equivalent pressure of the shock wave with different colored bounding spheres and the damaged area of the large-space complex building structure with different color blocks at different time nodes.

[0007] Another example is to refer to "Shock Wave Detection and Visualization in a Three-Dimensional Explosion Field", Zhang Wenyao et al., Journal of Beijing Institute of Technology, Vol. 34, Supplement 1, July 2014. This paper discloses a volume rendering process based on ray casting and a visualization acceleration processing method using techniques such as look-up tables and bounding boxes.

[0008] Another example is the invention patent application with the patent application number CN201810354700.9, which discloses a three-dimensional grid separation simulation system and method based on text mining. This three-dimensional grid separation simulation method uses spatial position for clustering to try to reduce the data calculation amount. However, the above grid simulation has a strong dependence on hardware resources in the visualization process (such as the graphics rendering or display process), so the computing power requirements for the computing device are relatively high.

[0009] 2) Using the material point method for simulation calculation

[0010] The material point method is a method used to simulate fluids and solids. In the material point method, material points are used to represent fluids and solids, and grids are used to represent the positions and velocities of the material points. Before the simulation begins, the material points need to be gridded, assigned to grid cells, and the mass and velocity of the material points in the grid cells need to be calculated. For example, for the visualization of the material point method, Huang Peng, Zhang Xiong, Ma Shang, et al. proposed a method for solving problems such as hypervelocity collision, penetration, and explosion using MPM3D software (see, Huang Peng, Zhang Xiong, Ma Shang, et al. Research on parallelization of three-dimensional explicit material point method based on OpenMP [J], Chinese Journal of Computational Mechanics, 010, 27(1):7. DOI:CNKI:SUN:JSJG.0.2010-01-004).

[0011] However, in the actual explosion simulation solution process, as the accuracy of the mass point increases, the size of the result file also increases rapidly, which also puts higher requirements on the hardware and software processing capabilities of the visualization process. At present, the visualization solution process of the complex explosion simulation process is: first use ANSYS software to model, divide the grid and generate a k file, then use the material point conversion program that matches the MPM3D software to convert the k file into a material point file that meets the requirements of MPM3D, and finally input the material point file into the MPM3D software to complete the subsequent visualization display.

[0012] In this process, in order to speed up the visualization process, clustering can be used to accelerate it. For example, before visualization rendering, clustering can be performed on the data result file to reduce the number of grid points in the data result file, thereby reducing the data processing pressure in the subsequent rendering process. However, this method also has certain defects, such as the quality of visualization is greatly affected, that is, the authenticity and reliability of the visualization results are reduced.

[0013] Therefore, there is an urgent need for a method that can accelerate the visualization of explosion damage simulation while ensuring the quality of visualization rendering. Summary of the invention

[0014] The present invention aims to provide a visualization acceleration method based on clustering in the post-processing process of explosion damage simulation, aiming at the problems of post-processing file reading and loading, data processing and analysis, and visualization rendering, so as to reduce the requirements of visualization rendering hardware and software and quickly evaluate the solution effect without affecting the visualization quality or accepting an acceptable loss of accuracy.

[0015] In order to solve the above-mentioned technical problems, the present invention specifically adopts the following technical solutions:

[0016] A clustering-based particle algorithm visualization acceleration method includes the following steps:

[0017] S101 Obtain a data result file, where the data result file includes: raw data for simulating the explosion damage process of an explosion object;

[0018] S102 Obtain the sampling scenario features of the data result file, and find a clustering undersampling method that matches the sampling scenario features in the clustering undersampling method library according to a preset matching rule; wherein, the sampling scenario features include one or more of the following: the type of the explosion object, the time period of the explosion damage process, the cohesion characteristics of the raw data, the amount of raw data, the attribute value of the raw data; the clustering undersampling method library includes one or more of the following methods: k-means clustering algorithm, density-based spatial clustering algorithm, comprehensive hierarchical clustering algorithm;

[0019] S103 Select a clustering setting value of a first clustering condition that matches the sampling scenario features and / or the clustering undersampling method from the first clustering condition database through the sampling scenario features and / or the clustering undersampling method matched in S102, and the first clustering condition is used to adjust the number of clusters in the clustering undersampling process;

[0020] S104 Select a second clustering condition input by the user through the clustering undersampling method, and obtain a clustering input value input by the user for the second clustering condition, and the second clustering condition is used to adjust the density of the clusters in the clustering undersampling process;

[0021] S105 Use the clustering setting value and the clustering input value to perform clustering undersampling on the visual physical features of the data result file through the corresponding clustering undersampling method.

[0022] In some embodiments, when the clustering undersampling method is the k-means clustering algorithm, the first clustering condition includes: the number of initial clustering centers, and the second clustering condition includes: the number of clusters (k value).

[0023] In some embodiments, when the clustering undersampling method is the density-based spatial clustering algorithm,

[0024] The first clustering condition includes: the minimum sampling value (min_samples); the second clustering condition includes: the sample distance of the cluster (eps).

[0025] In some embodiments, when the clustering undersampling method is the comprehensive hierarchical clustering algorithm, the first clustering condition includes: the maximum entry tree (Branching Factor) of the leaf nodes of the clustering feature tree;

[0026] The second clustering condition includes: the size of the clustering feature tree (Threshold).

[0027] In some embodiments, the visualized physical features include one or more of the following: material (type of material); damage (degree of damage); pressure.

[0028] In some embodiments, before S105, the method further includes the step of:

[0029] obtaining the visualization requirements associated with the data result file, where the visualization requirements include one or more of the following: shock wave visualization, damage degree visualization;

[0030] selecting, from the clustered physical features in the data result file, the clustered physical features that match the visualization requirements according to the visualization requirements, and marking the selected clustered physical features as visualized physical features.

[0031] In some embodiments, the method further includes the steps of:

[0032] S106 obtaining the sampling data features from the sampling results of the clustered downsampling, where the sampling data features include: the position of the cluster center, and the number represented by this cluster;

[0033] S107 calculating, based on the sampling data features, the discard point features in the process of clustered downsampling, where the discard point features include: the number of discard points, and / or the distribution of the discard points;

[0034] S108 determining whether the sampling data features match the original data features of the data result file. If so, performing visualization calculation according to the sampling results; if not, correcting the clustered input value, and performing step S105 according to the corrected clustered input value.

[0035] In a second aspect of the present invention, there is also provided a visualization acceleration system for a clustering-based particle-like algorithm, including:

[0036] a data acquisition module configured to acquire a data result file, where the data result file includes: original data for simulating the explosion damage process of an explosion object;

[0037] a downsampling method matching module configured to acquire the sampling scenario features of the data result file, and find a clustered downsampling method that matches the sampling scenario features in a clustered downsampling method library according to a preset matching rule; where the sampling scenario features include one or more of the following:

[0038] The second aspect of the present invention lies in that there is also provided a visualization acceleration system for a clustering-based particle-like algorithm, including:

[0039] including:

[0040] a data acquisition module configured to acquire a data result file, the data result file including: original data for simulating the explosion damage process of an explosion object;

[0041] a downsampling method matching module configured to acquire the sampling scenario features of the data result file, and find a clustered downsampling method that matches the sampling scenario features in a clustered downsampling method library according to a preset matching rule; wherein, the sampling scenario features include one or more of the following:

[0042] The type of the explosion object, the time period of the explosion damage process, the cohesion characteristics of the original data, the amount of the original data, and the attribute values of the original data; The clustering downsampling method library includes one or more of the following methods: k-means clustering algorithm, density-based spatial clustering algorithm, and comprehensive hierarchical clustering algorithm;

[0043] The first clustering condition setting module is configured to select, from the first clustering condition database, a clustering setting value of the first clustering condition that matches the sampling scenario feature and / or the clustering downsampling method through the clustering downsampling method matched in the sampling scenario feature and / or sampling method matching module, where the first clustering condition is used to adjust the number of clusters in the clustering downsampling process;

[0044] The second clustering condition setting module is configured to select, through the clustering downsampling method, a second clustering condition input by the user and obtain a clustering input value input by the user for the second clustering condition, where the second clustering condition is used to adjust the density of the clusters in the clustering downsampling process;

[0045] The downsampling module is configured to perform clustering downsampling on the visual physical features of the data result file by using the clustering setting value and the clustering input value through the corresponding clustering downsampling method.

[0046] In some embodiments, when the clustering downsampling method is the k-means clustering algorithm, the first clustering condition includes: the number of initial clustering centers, and the second clustering condition includes: the number of clusters;

[0047] When the clustering downsampling method is the density-based spatial clustering algorithm, the first clustering condition includes: the minimum sampling value; the second clustering condition includes: the sample distance of the clusters;

[0048] When the clustering downsampling method is the comprehensive hierarchical clustering algorithm, the first clustering condition includes:

[0049] The maximum entry tree of the leaf nodes of the clustering feature tree; the second clustering condition includes: the size of the clustering feature tree.

[0050] In some embodiments, the visual physical features include one or more of the following: material;

[0051] damage; pressure.

[0052] Beneficial technical effects:

[0053] In order to ensure the reliability of the visualization effect and the solution efficiency while accelerating the visualization processing process, on the one hand, the present invention selects a limited type of sampling scenario features as the selection conditions for clustering downsampling, so as to quickly match the clustering downsampling method that meets the visualization requirements in the limited method library; on the other hand, for each clustering downsampling method, the clustering conditions in the clustering process are divided into preset values and real-time input values, so as to minimize the call startup time of different clustering downsampling methods.

[0054] In other words, for the visualization processing of explosion problems, on the one hand, the clustering downsampling method and the real-time input value of clustering downsampling are used as selective adjustment factors to meet the real explosion solution requirements, and on the other hand, the types of typical sampling scenario features, clustering downsampling methods, and real-time input values of clustering downsampling are limitedly selected, thus providing a user with a clustering downsampling method with good versatility and autonomy, and the pre-setting or enabling time of this method is relatively short.

[0055] In summary, the present invention provides a semi-automatic visualization acceleration method that combines standardized automatic setting and manual adjustment. It uses a standardized program to automatically set the clustering downsampling method, the first clustering condition, the visualization physical characteristics, etc., and selects the second clustering condition as the manual adjustment port to superimpose the advantages of manual adjustment to coordinate the differences or contradictions between the standardized program and the actual application scenario (for example, the data characteristics of VTU files, the visualization requirements of different research tasks). Thus, this semi-automatic visualization acceleration method can not only have the benefits of flexibility and authenticity brought by manual adjustment, but also greatly reduce the difficulty of manual intervention (for example, avoid staff spending too much energy on parameter selection or setting, or reduce the professional level requirements for staff).

[0056] Moreover, this semi-automatic visualization acceleration method that combines standardized automatic setting and manual adjustment can not only quickly reduce the data dimension of large-scale data sets, reduce the number of data points required for visualization, and thus reduce the computational and rendering burden of visualization. At the same time, it can also maintain the key features of the data after the data dimension is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts do not necessarily draw according to the actual scale. Obviously, the following-described drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0058] Figure 1a It is a schematic flow chart of a visualization acceleration method for a clustering-based particle algorithm in an exemplary embodiment of the present invention;

[0059] Figure 1b It is a schematic diagram of the VTU file structure;

[0060] Figure 2 It is a schematic diagram of the module structure of a visualization acceleration system for a clustering-based particle algorithm in an exemplary embodiment of the present invention;

[0061] Figure 3 It is a schematic diagram of the effect before downsampling for high-speed collision problem verification;

[0062] Figure 4 It is a schematic diagram of the effect after downsampling for high-speed collision problem verification;

[0063] Figure 5 It is a schematic diagram of the effect before downsampling for penetration explosion problem verification;

[0064] Figure 6 It is a schematic diagram of the effect after downsampling for penetration explosion problem verification. Detailed implementation manners

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0066] In this article, suffixes such as "module", "component", or "unit" used to represent elements are only for the convenience of explaining the present invention, and they have no specific meaning in themselves. Therefore, "module", "component", or "unit" can be used interchangeably.

[0067] In this article, terms such as "upper", "lower", "inner", "outer", "front", "rear", "one end", "the other end", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation to the present invention. In addition, terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0068] In this text, unless otherwise clearly stipulated and defined, terms such as "installation", "equipped with", "connection", etc.,

[0069] should be understood in a broad sense. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, or an indirect connection through an intermediate medium, and can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0070] In this text, "and / or" includes any and all combinations of one or more of the listed related items.

[0071] In this text, "a plurality of" means two or more, that is, it includes two, three, four, five, etc.

[0072] It should be noted that in this text, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article or device. Without more limitations, an element defined by the statement "comprising one..." does not exclude the existence of another identical element in the process, method, article or device comprising that element.

[0073] As used in this specification, the term "about" typically represents + / - 5% of the stated value,

[0074] more typically + / - 4% of the stated value, more typically + / - 3% of the stated value, more typically + / - 2% of the stated value, even more typically + / - 1% of the stated value, and even more typically

[0075] + / - 0.5% of the stated value.

[0076] In this specification, certain embodiments may be disclosed in a format within a certain range. It should be understood that this kind of description "within a certain range" is only for convenience and brevity, and should not be construed as a rigid limitation on the disclosed range. Therefore, the description of the range should be considered to have specifically disclosed all possible sub-ranges and individual numerical values within this range. For example, the description of the range should be regarded as having specifically disclosed sub-ranges such as from 1 to 3, from 1 to 4, from 1 to 5, from 2 to 4, from 2 to 6, from 3 to 6, etc., and individual numbers within this range, such as 1,

[0077] 2, 3, 4, 5 and 6. The above rules apply regardless of the breadth of the range.

[0078] In this article, "data result file" refers to the result file in the actual explosion simulation solution process. For example, the data result file can be a large volume result data (VTU) file. For example, the data result file can be a file in ASCII and binary formats.

[0079] Herein, the "attribute value of data" may be a numerical value associated with the visualization of data. For example, for a VTU file, its attribute value may be a physical quantity such as "mate (material)" or "damg (damage)".

[0080] In this article, the "second clustering condition" is also referred to as a custom parameter group, which includes one or more parameters that can be defined and input by the user. Through the setting of the custom parameter group, different views and details of the exploration data can be presented to the user.

[0081] When the type of explosion object is relatively complex or the problem is difficult to solve, for example, in a simulation scene with 1.3 billion particles, the result file (such as VTU file) for each time step often reaches about 50G. For large VTU files, visual rendering has problems such as file reading and loading time-consuming, data processing and analysis require more complex and efficient algorithms and tools, and graphics rendering and display rely on hardware resources.

[0082] In order to solve the problems of file reading and loading, data processing and analysis, and visualization rendering faced by large volume data result (VTU) files during visualization, the present invention provides a visualization acceleration method and system for explosion damage simulation based on clustering.

[0083] The system uses a semi-automatic processing mode that combines standardized program processing with manual adjustment. It reduces the data dimension of unstructured grid data without affecting the visualization quality or acceptable loss of accuracy, while retaining the detailed features of the original data as much as possible. In this way, it can reduce the hardware requirements for visualization rendering, quickly evaluate the solution effect, and meet the visualization quality requirements.

[0084] Embodiment 1

[0085] In order to solve the problems of reading, writing, loading, data processing and analysis, and visualization rendering of post-processing files (i.e., data result files) of explosion damage simulation, the present invention provides a clustering-based particle algorithm visualization acceleration method to reduce the hardware requirements for visualization rendering of complex explosion objects (such as large buildings, complex mountain scenes, tunnel scenes, etc.) without affecting the quality of visualization or on the basis of acceptable precision loss, and improve the efficiency of visualization rendering.

[0086] likeFigure 1a As shown in the figure, the present invention provides a visualization acceleration method for a particle-based algorithm based on clustering (which is also equivalent to providing a post-processing method), including the steps:

[0087] S101 Obtain a data result file, where the data result file includes: raw data for simulating the explosion damage process of an explosion object.

[0088] In some embodiments, the data result file is a data result file during the explosion simulation of a building and its surrounding scenes, mountain scenes, etc., such as a VTU file.

[0089] For example, in some embodiments, by directly inputting the data result file into an existing visualization software (e.g., MPM3D), the visualization presentation of the explosion damage of buildings and mountains can be completed.

[0090] For example, in some embodiments, the raw data in the data result file can be particle data (preferably material points).

[0091] S102 Obtain the sampling scene features of the data result file, and find a clustering downsampling method that matches the sampling scene features in the clustering downsampling method library according to a preset matching rule; where the sampling scene features include one or more of the following: the type of explosion object, the time period of the explosion damage process, the cohesion characteristics of the raw data, the amount of raw data, the attribute values of the raw data; the clustering downsampling method library includes one or more of the following methods: k-means clustering algorithm, density-based spatial clustering algorithm, comprehensive hierarchical clustering algorithm.

[0092] In some embodiments, according to the data result characteristics of complex explosion objects, first select limited factors such as the type of explosion object, the time period of the explosion damage process, the cohesion characteristics of the raw data, the amount of raw data, the attribute values of the raw data, etc. as typical sampling scene features. And correspondingly select 3 clustering methods such as the k-means clustering algorithm, the density-based spatial clustering algorithm, and the comprehensive hierarchical clustering algorithm as the main clustering means to meet the clustering processing requirements for different typical sampling scene features.

[0093] In some embodiments, the matching rule can be freely set by the user according to historical work experience.

[0094] For example, in some embodiments, when the explosion object is a complex scene such as a building complex or a mountain, the corresponding clustering downsampling method preferably matches the density-based spatial clustering algorithm (such as the DBSCAN clustering method) or the comprehensive hierarchical clustering algorithm (such as the BIRCH clustering method). When the explosion object is a small single-family building, the corresponding clustering downsampling method is the k-means clustering algorithm.

[0095] For example, in some embodiments, when the amount of original data in the data result file is large and the cohesion characteristic of the original data is that the number of features is small, the BIRCH clustering method is preferably used.

[0096] For example, in some embodiments, when the cohesion characteristic of the original data in the data result file is that the cluster structure is not obvious, or the aggregation size is unbalanced, and the type of explosion object is a complex object such as a building complex or a mountain, or the time period of the explosion damage process is the initial stage, the DBSCAN clustering method is preferably used.

[0097] For example, in some embodiments, the clustering downsampling methods in the clustering downsampling method library are associated with corresponding matching tags, and the matching tags include: at least one sampling scenario feature information.

[0098] In some embodiments, when two or more clustering downsampling methods are matched,

[0099] the preferred clustering downsampling method can be selected according to the application priority of the clustering downsampling method; or,

[0100] a prompt signal can be input to the user to actively select a suitable clustering downsampling method by the user.

[0101] S103 selects the clustering setting value of the first clustering condition that matches the sampling scenario feature and / or the clustering downsampling method from the first clustering condition database through the sampling scenario feature and / or the clustering downsampling method matched in S102. The first clustering condition is used to adjust the number of clusters in the clustering downsampling process.

[0102] S104 selects the second clustering condition input by the user through the clustering downsampling method and obtains the clustering input value input by the user for the second clustering condition. The second clustering condition is used to adjust the density of clusters in the clustering downsampling process.

[0103] S105 uses the clustering setting value and the clustering input value to perform clustering downsampling on the visual physical characteristics of the data result file through the corresponding clustering downsampling method.

[0104] In the visualization acceleration method of this embodiment, on the one hand, the applicable clustering downsampling method can be quickly matched from the limited method database by using the preset matching rule, thus providing a standardized processing procedure for typical clustering downsampling scenarios and visualization simulation requirements; on the other hand, this method also designs limited artificial input data (such as the clustering input value manually set by the user), thereby enhancing the flexible adaptability between the standardized processing procedure and different typical downsampling scenarios or visualization simulation requirements, that is, enhancing the adaptability between the standardized processing procedure and the actual application scenario.

[0105] Therefore, the present invention actually provides a standardized clustering downsampling method, which can not only use a standardization processing program to quickly and massively reduce the data dimension of the data result file and reduce the pressure of visual rendering, but also cooperate with a typical clustering downsampling method and manually input data to limit the loss of visual quality within an acceptable range of typical application requirements.

[0106] For example, in some embodiments, before inputting the data result file into the visual simulation software

[0107] - MPM3D software, the above acceleration method can be used to reduce the data dimension of the original data result file on the one hand, thereby reducing the data processing pressure of the MPM3D software and the computing device hardware. At the same time, the coordination between the standardization processing and the manual borrowing can ensure that the data after dimension reduction still has reliable data characteristics, so as to still maintain good detail effects after the material point simulation (or, after visualization).

[0108] In some embodiments, when two or more clustering downsampling methods are matched, the recommended sampling method can also be selected according to the application priority of the clustering downsampling method.

[0109] For example, in some embodiments, the clustering downsampling method is also marked with an application priority. For example,

[0110] In some specific embodiments, the k-means clustering algorithm is marked with a first application priority, the density-based spatial clustering algorithm is marked with a second application priority, and the comprehensive hierarchical clustering algorithm is marked with a third application priority.

[0111] It can be understood that the above setting order of the priorities is only for exemplary illustration, and the present invention does not make any limitations.

[0112] Again, for example, in some embodiments, for a scenario with relatively more scenario features, the clustering downsampling method can also be matched according to the matching priorities of different scenario features to be sampled and in the recommended order.

[0113] For example, in some embodiments, at least one sampling scenario feature is marked with a corresponding matching priority, and when the number of the obtained sampling scenario features is two or more, the corresponding clustering downsampling methods are matched in turn according to the matching priorities of the sampling scenario features. And when two or more clustering downsampling methods are matched, the clustering downsampling method with the highest priority can be selected according to the matching priority. Or, the user can also manually select the corresponding clustering downsampling method according to the matching result of the clustering downsampling method.

[0114] In some embodiments, when the clustering downsampling method is the k-means clustering algorithm, the first clustering condition includes: the number of initial clustering centers, and the second clustering condition includes: the number of clusters (such as the k value).

[0115] In some embodiments, when the clustering downsampling method is the density-based spatial clustering algorithm,

[0116] the first clustering condition includes: the minimum sampling value (such as min_samples); the second clustering

[0117] condition includes: the sample distance of the clustering (such as eps);

[0118] In some embodiments, when the clustering downsampling method is the comprehensive hierarchical clustering algorithm, the first clustering condition includes: the maximum entry tree of the leaf nodes of the clustering feature tree (such as BranchingFactor); the second clustering condition includes: the size of the clustering feature tree (such as Threshold).

[0119] In some embodiments, the visualized physical features include one or more of the following:

[0120] (i) material, that is, the type of material, which can characterize the material properties of points (which can also be called mass points, particles, grid points, etc.), such as air, water, steel, concrete, etc.

[0121] For example, in some embodiments, material can be the number of the material.

[0122] (ii) damage, that is, the degree of damage, which can characterize the degree of damage of points.

[0123] For example, in some embodiments, the degree of damage in concrete is the sum of the damage caused by equivalent plastic strain and the damage caused by plastic volumetric strain.

[0124] (iii) pressure, that is, the pressure, which can characterize the pressure value of points.

[0125] For example, in some embodiments, the pressure can be the gas pressure representing air mass points, or the opposite of the tensile force in solid mass points.

[0126] For example, in some embodiments, when performing explosion simulation modeling, when the complexity of constructing explosion objects such as buildings or the problem domain is relatively high, it is preferably to select the parameter combination formed by material, damage, and pressure as the visual physical features (i.e., clustering objects) to ensure the visualization effect to the greatest extent. In other words, selecting material, damage, and pressure as the core clustering objects can largely reduce the adverse impact of data dimension on visualization quality.

[0127] In some embodiments, before S105, the method further includes the steps of:

[0128] Obtaining the visualization requirements associated with the data result file, where the visualization requirements include one or more of the following: shock wave visualization, damage degree visualization;

[0129] Selecting the clustering physical features that match the visualization requirements from the clustering physical features in the data result file according to the visualization requirements, and marking the selected clustering physical features as visual physical features.

[0130] In this embodiment, the user can change the visualization result by selecting the clustering clusters of interest, or adjusting the clustering parameters, or selecting the clustering objects to focus on. This interactivity and controllability enable the user to customize and explore different views and details of the data.

[0131] For example, in some embodiments, when the original data volume of the data result file is relatively large, or when facing some unconventional explosion solving tasks, the necessary visual physical features can be screened before clustering processing.

[0132] For example, in some embodiments, for the shock wave visualization requirement, at least pressure needs to be selected as the necessary visual physical feature.

[0133] For another example, in some embodiments, for the damage degree visualization requirement, at least material and damage are selected as the necessary visual physical features.

[0134] For example, in some embodiments, the pre-stored clustering physical features are associated with matching labels, and the matching labels include: at least one visualization requirement.

[0135] For example, in some embodiments, the automatically matched visual physical features and the self-selected visual physical features automatically input by the user can be combined as the clustering objects to complete the clustering downsampling process.

[0136] For example, in some embodiments, the pre-stored clustering physical features are associated with matching labels, and the matching labels include: at least one visualization requirement.

[0137] For example, in some embodiments, the automatically matched visual physical features and the self-selected visual physical features automatically input by the user can be combined as the clustering objects to complete the clustering downsampling process.

[0138] In some embodiments, it further includes the steps:

[0139] S106 Obtain sampling data features from the sampling results of clustering downsampling, and the sampling data features

[0140] include: the position of the cluster center, and the number represented by this cluster;

[0141] S107 Calculate the discard point features in the clustering downsampling process through the sampling data features, and the discard point features

[0142] include: the number of discard points, and / or the distribution of discard points;

[0143] S108 Determine whether the sampling data features (such as discard point features) match the original data features of the data result file. If so, perform visualization calculation according to the sampling results; if not,

[0144] correct the clustering input value, and perform step S105 according to the corrected clustering input value.

[0145] In some embodiments, the original data features can be the clustering distribution features of the data in the data result file, or the preliminary clustering quantity.

[0146] Alternatively, in some embodiments, the original data features can also be custom parameter thresholds according to the actual situation.

[0147] Preferably, in some embodiments, when the discard point features do not match the original data features,

[0148] send a corresponding prompt signal to the user, update the clustering input value in the previous clustering process according to the newly input clustering input value by the user, and then perform clustering processing again.

[0149] In order to further make the visualization effect closer to the actual business requirements, in this embodiment, a method of multiple loop adjustments can also be adopted to continuously optimize the downsampling clustering results. Specifically, before performing visualization calculation, this embodiment selects the discard point features as the typical judgment criterion for sampling quality, so as to assist the user to quickly complete the numerical judgment and adjustment in the multiple loop process through self-defined matching and manual intervention. Therefore, the verification method in this embodiment can pre-evaluate the visualization quality in a way with relatively limited data processing volume before performing the complete visualization task, so as to reduce or avoid the consumption time of visualization calculation to a certain extent (such as avoiding resource consumption due to repeated calculation).

[0150] For example, in some embodiments, the evaluation conditions or criteria can be determined according to the characteristics of different downsampling methods. For example, according to the selection of the initial points of the k-means method and different purposes such as determining the min_samples parameter of dbscan, the evaluation is carried out separately.

[0151] For example, in some embodiments, the setting of the evaluation conditions or criteria can also be determined in combination with the shape of the model, the effects of explosion and penetration, etc.

[0152] For example, in some embodiments, during the downsampling process, the positions of the cluster centers and the number of representatives of each cluster can be printed first, and then the number of discarded points can be obtained; the staff can roughly estimate the correctness and effectiveness of the downsampling based on their understanding of the model.

[0153] Alternatively, in some other embodiments, the visualization effects of the post-processing file and the source file (i.e., the data result file) after downsampling can also be compared and evaluated, and then the sizes of the key parameters can be adjusted.

[0154] To verify the reliability of the visualization method provided by the present invention, application examples of high-speed collision problems and penetration explosion problems are also provided in this article to verify the reliability of the above acceleration method.

[0155] Specifically, the experimental results of the clustering downsampling experiment on the data of two experiments are as follows:

[0156] Experimental results of high-speed collision problem verification

[0157] See

[0158] See Figure 3 and Figure 4 , which are respectively the schematic diagrams of the visualization effects of the original data and the clustering downsampling data of the data result file of the high-speed collision problem verification experiment. Through the comparison of Figure 3 and Figure 4 , it can be seen that after the visualization acceleration processing of the present invention, the similarity between the obtained clustering downsampling data result and the original data is relatively high, and good detail features are still maintained in the visualization effect presentation.

[0159] Experimental results of penetration explosion problem verification

[0160] See Figure 5 and Figure 6 as shown. Figure 5 and Figure 6 are respectively the schematic diagrams of the visualization effects of the original data and the clustering downsampling data of the data result file of the penetration explosion problem verification experiment. From the comparison of Figure 5 and Figure 6From the comparison of the effects, it can be seen that even when the downsampling ratio is relatively large (that is, when the degree of data dimension reduction is relatively high), Figure 6 the visualization effect in

[0161] Figure 6 still retains the main core details. In other words,

[0162] the loss of visualization quality in

[0163] is controlled within an acceptable range for users.

[0164] Embodiment 2

[0165] Referring to Figure 2 shown, corresponding to the visualization acceleration method in the above embodiment, the present invention also provides

[0166] a visualization acceleration system for a particle-based algorithm based on clustering, including:

[0167] A data acquisition module 11, configured to acquire a data result file, where the data result file includes: original data for simulating the explosion damage process of an explosion object;

[0168] A downsampling method matching module 12, configured to acquire the sampling scenario features of the data result file, and find a clustering downsampling method that matches the sampling scenario features in a clustering downsampling method library according to a preset matching rule; wherein, the sampling scenario features include one or more of the following: the type of the explosion object, the time period of the explosion damage process, the cohesion characteristics of the original data, the amount of the original data, the attribute value of the original data; the clustering downsampling method library includes one or more of the following methods: k-means clustering algorithm, density-based spatial clustering algorithm, comprehensive hierarchical clustering algorithm;

[0169] A first clustering condition setting module 13, configured to select a clustering setting value of a first clustering condition that matches the sampling scenario features and / or the clustering downsampling method from a first clustering condition database through the sampling scenario features and / or the clustering downsampling method matched in the sampling method matching module, where the first clustering condition is used to adjust the number of clusters in the clustering downsampling process;

[0170] The second clustering condition setting module 14 is configured to select the second clustering condition input by the user through the clustering downsampling method and obtain the clustering input value input by the user for the second clustering condition, where the second clustering condition is used to adjust the density of the clusters in the clustering downsampling process;

[0171] The downsampling module 15 is configured to use the clustering setting value and the clustering input value to perform clustering downsampling on the visual physical features of the data result file through the corresponding clustering downsampling method.

[0172] In some embodiments, when the clustering downsampling method is the k-means clustering algorithm, the first clustering condition includes: the number of initial cluster centers, and the second clustering condition includes: the number of clusters.

[0173] In some embodiments, when the clustering downsampling method is the density-based spatial clustering algorithm,

[0174] The first clustering condition includes: the minimum sampling value; the second clustering condition includes: the sample distance of the clusters.

[0175] In some embodiments, when the clustering downsampling method is the comprehensive hierarchical clustering algorithm, the first clustering condition includes: the maximum entry tree of the leaf nodes of the clustering feature tree; the second clustering condition includes: the size of the clustering feature tree.

[0176] In some embodiments, the visual physical features include one or more of the following: material;

[0177] damage; pressure.

[0178] In some embodiments, the system further includes:

[0179] The visualization requirement acquisition module 16 is configured to acquire the visualization requirements associated with the data result file, and the visualization requirements include one or more of the following: shock wave visualization, damage degree visualization;

[0180] The visual physical feature selection module 17 is configured to select the clustering physical features that match the visualization requirements from the clustering physical features in the data result file according to the visualization requirements, and mark the selected clustering physical features as visual physical features.

[0181] In some embodiments, the system further includes:

[0182] The sampled data feature acquisition module 18 is configured to acquire the sampled data features from the sampling results of the clustering downsampling, and the sampled data features include: the position of the cluster center, and the representative of this cluster

[0183] The quantity of;

[0184] The discard point feature acquisition module 19 is configured to calculate the discard point features during the clustering downsampling process through the sampled data features, and the discard point features include: the number of discard points, and / or the distribution of discard points;

[0185] The verification module 20 is configured to determine whether the sampled data features (such as discard point features) match the original data features of the data result file. If so, perform visualization calculation according to the sampling result; if not, correct the clustering input value and re-enter the corrected clustering input value into the downsampling module 15.

[0186] It can be understood that the visualization acceleration system in this embodiment can also execute the methods or steps in any of the above embodiments, which will not be elaborated here.

[0187] To more clearly illustrate the technical solutions and beneficial technical effects of the present invention, the following takes the VTU file

[0188] as an example for exemplary processing description:

[0189] The VTU file format is an open-source data storage format for storing and displaying scientific data, and its format can represent various dataset types, such as polygons, scalars, vectors, tensors, textures, etc. VTU

[0190] The file format has two modes: ASCII and binary.

[0191] And in the ASCII mode, the structure of the VTU file is as Figure 1b shown.

[0192] Among them, the first line of the VTU file is usually: #vtk DataFile Version x.x / / File format

[0193] Version;

[0194] The second line of the VTU file is usually: Any string / / File title;

[0195] The third line of the VTU file is usually: ASCII or BINARY / / Data mode;

[0196] The fourth line of the VTU file is usually: DATASET dataset_type / / Dataset type;

[0197] The subsequent lines of the VTU file are usually: Specific data content, including points, cells, attributes, etc.;

[0198] To cooperate with the post - processing method provided by the present invention, preferably the Python language and the VTK

[0199] module are used to access the data in the VTU file. Specifically, the exemplary access process is as follows:

[0200] 1) Create a vtkUnstructuredGridReader object to read the VTU file;

[0201] 2) Call the SetFileName method to set the name of the VTU file to be read;

[0202] 3) Call methods such as ReadAllScalarsOn, ReadAllVectorsOn, ReadAllTensorsOn

[0203] to set the data types to be read;

[0204] 4) Call the Update method to update the reader status;

[0205] 5) Call the GetOutput method to obtain the data object of the VTU file, that is, unstructuredgrid;

[0206] 6) Call the GetNumberOfPoints and GetNumberOfCells methods to obtain the number of points and cells

[0207] ;

[0208] 7) Call the GetPointData or GetCellData method to obtain the data sets of points or cells;

[0209] 8) Call methods such as GetScalars, GetVectors, GetTensors to obtain the specific data

[0210] arrays, such as potential or velocity;

[0211] 9) Use the numpy_support module to convert the vtk array to a numpy array for convenient subsequent processing.

[0212] After the data in the VTU file is read, a clustering down - sampling method and a clustering setting value adapted thereto can be further selected.

[0213] Moreover, to ensure that the topology and shape features of the mesh remain unchanged after dimensionality reduction of large-volume VTU files, defects such as holes, cracks, or distortions are avoided. Preferably, according to the preselected typical downsampling scenario features, the corresponding clustering downsampling method is quickly matched through a standardization program, and the preset first clustering condition is correspondingly selected to help the user quickly find a clustering downsampling method and clustering condition with better adaptability. Also, for different clustering downsampling methods, a customizable second clustering condition input port is provided to the user to further retain the details and differential features of the data on the basis of reducing data loss.

[0214] Preferably, in some embodiments, according to different scenarios of downsampling large-volume VTU files, three targeted clustering methods are selected to meet the downsampling requirements in actual work.

[0215] For example, in some embodiments, for scenarios where the cohesion characteristics of the data result file are spherical, or the clustering distribution is relatively regular, or when the problem domain constructed by the explosion object or damage process is relatively simple, the k-means clustering method is preferably used.

[0216] The exemplary process of processing the data result file using the k-means method is as follows:

[0217] Select the clustering setting value of the first clustering condition matching k-means - the number of initial clustering centers, and the number of clusters (such as the K value) manually input by the user to perform clustering downsampling on the data result file.

[0218] In some embodiments, the clustering setting value can be an empirical value set by the user according to historical experience. For example, the selection of the number of initial clustering centers can be determined by the visualization physical quantity to be displayed.

[0219] Or, the number of initial clustering centers can also be calculated by the k-means++ algorithm.

[0220] Or, in some embodiments, the clustering setting value can also be a historical setting value, such as the clustering setting value used in the historical clustering process of data result files with the same or similar downsampling scenarios.

[0221] Again, for example, in some embodiments, when the explosion object or the scene environment where the explosion object is located is relatively complex. Or, when there are many noise points (i.e., points that do not belong to any cluster) in the original data, the DBSCAN clustering method is preferably used as the clustering downsampling method.

[0222] The exemplary process of processing the data result file using the DBSCAN clustering method is as follows:

[0223] First, organize and summarize the data in the given VTU file to form a physical property dataset. In the given dataset, according to the density of other data points around each data point, the data points are divided into core points, boundary points, and noise points. A core point is a data point with a sufficient number of other data points within a certain radius; a boundary point is a data point that does not meet the requirements of a core point but is within the radius of a certain core point;

[0224] A noise point is a point that does not meet any conditions.

[0225] In some embodiments, for the selected DBSCAN clustering method and the attribute values of the data, the corresponding first clustering condition - min_samples is selected. And eps is selected as the second clustering condition to serve as an adjustment port for the user to flexibly adjust.

[0226] In some embodiments, the default setting value of min_samples is 30.

[0227] Alternatively, in some other embodiments, different setting values of min_samples can also be matched according to the visualized data density, outliers, etc.

[0228] Among them, the eps parameter is a radius parameter used to determine the neighborhood of a point. If the distance between two points is less than eps, then these two points are considered density-connected points. The min_samples parameter is the minimum number of samples parameter used to determine a core point. If there are at least

[0229] min_samples points in the neighborhood of a point, then this point is considered a core point.

[0230] In some embodiments, when the requirement for visualization quality is high, the value of eps can also be adjusted.

[0231] Its exemplary adjustment process is as follows:

[0232] 1) Select an eps value, and then use the DBSCAN algorithm to cluster the dataset.

[0233] 2) Evaluate the clustering result. If the clustering result is not ideal, increase the eps value and re-run

[0234] the DBSCAN algorithm.

[0235] Repeat step 2) until the best clustering result is obtained.

[0236] In some embodiments, in order to further reduce the loss of visualization quality, it is also possible to

[0237] Adjust the min_samples value. An exemplary adjustment process is as follows:

[0238] 1) Select a min_samples value, and then use the DBSCAN algorithm to cluster the dataset.

[0239] 2) Evaluate the clustering result. If the clustering result is not satisfactory, increase the min_samples value and re-run the DBSCAN algorithm.

[0240] Repeat step 2) until the best clustering result is obtained.

[0241] In this embodiment, by selecting and adjusting the first and second clustering conditions, on the one hand, incorrect assignment of noise points can be avoided, and on the other hand, improper overlap between clusters can also be avoided.

[0242] For another example, in some embodiments, when the original data is massive and the number of features is relatively small, it is preferably to use the BIRCH clustering method as the clustering downsampling method. Correspondingly, set the Branching Factor as the first clustering condition, where the Branching Factor is the maximum number of entries that each leaf node in the CF-Tree can contain.

[0243] In some embodiments, the default setting value of the Branching Factor is 200.

[0244] Or, in some other embodiments, the default setting value of the Branching Factor is between 50 and 300. Specifically, in the actual application process, either match the corresponding value according to the data characteristics of the data result file, or freely select by the user within the limited range.

[0245] Among them, the BIRCH clustering method uses a data structure called a CF tree to represent the dataset. This method mainly includes two core processes: constructing a CF tree and extracting clustering results from the CF tree.

[0246] In the process of constructing the CF tree, the BIRCH clustering method uses a data structure called a BIRCH node to represent the CF tree. A BIRCH node contains three parts: a clustering feature vector, a clustering radius, and a pointer to a child node. The BIRCH clustering method uses a statistic called CF (clustering feature) to calculate the clustering feature vector and the clustering radius. The CF statistic includes the number of samples, the sum of samples, the sum of squares of samples, and the covariance matrix of samples.

[0247] In the process of extracting clustering results from the CF tree, the BIRCH clustering method uses an algorithm called CF clustering to extract the clustering results. The CF clustering algorithm uses a threshold parameter to determine the clustering radius. If the distance between two nodes is less than the clustering radius, these two nodes are considered density-connected nodes.

[0248] The BIRCH clustering method has two parameters: threshold and n_clusters.

[0249] The threshold parameter is the threshold parameter in the BIRCH algorithm. It is used to control the size of the CF tree.

[0250] If the distance between the CF statistic of a node and the clustering feature vector of the current node is less than the threshold, the node is added to the children of the current node. If the distance between the CF statistic of a node and the clustering feature vector of the current node is greater than the threshold, the node is added to the CF tree as a new node. In this embodiment, the threshold is set as a clustering input value that can be flexibly adjusted by the user.

[0251] The n_clusters parameter is the number-of-clusters parameter in the BIRCH algorithm. It is used to specify the number of clusters for clustering. If the value of the n_clusters parameter is None, the BIRCH clustering method will automatically determine the number of clusters for clustering. If the value of the n_clusters parameter is an integer, the BIRCH clustering method will cluster the dataset into the specified number of clusters.

[0252] It should be noted that in this article, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including that element.

[0253] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described example methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions for causing a computer terminal (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0254] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the purpose of the present invention and the scope protected by the claims. All of these fall within the protection scope of the present invention.

Claims

1. A visualization acceleration method for particle-based algorithms based on clustering, characterized in that, it includes the steps: S101 Obtain a data result file, where the data result file includes: raw data for simulating the explosion damage process of an explosion object; S102 Obtain the sampling scenario features of the data result file, and find a clustering downsampling method that matches the sampling scenario features in the clustering downsampling method library according to a preset matching rule; wherein, the sampling scenario features include one or more of the following: the type of the explosion object, the time period of the explosion damage process, the cohesion characteristics of the raw data, the amount of raw data, the attribute values of the raw data; the clustering downsampling method library includes one or more of the following methods: k-means clustering algorithm, density-based spatial clustering algorithm, comprehensive hierarchical clustering algorithm; S103 Select a clustering setting value of the first clustering condition that matches the sampling scenario features and / or the clustering downsampling method from the first clustering condition database through the sampling scenario features and / or the clustering downsampling method matched in S102, and the first clustering condition is used to adjust the number of clusters in the clustering downsampling process; S104 Select a second clustering condition input by the user through the clustering downsampling method, and obtain a clustering input value input by the user for the second clustering condition, and the second clustering condition is used to adjust the density of clusters in the clustering downsampling process; S105 Use the clustering setting value and the clustering input value to perform clustering downsampling on the visualization physical features of the data result file through the corresponding clustering downsampling method; wherein, the second clustering condition serves as an artificial adjustment port, and the first clustering condition and the visualization physical features are automatically set through a standardization program.

2. A visualization acceleration method for particle-based algorithms based on clustering according to claim 1, characterized in that, when the clustering downsampling method is the k-means clustering algorithm, the first clustering condition includes: the number of initial cluster centers, and the second clustering condition includes: the number of clusters.

3. A visualization acceleration method for particle-based algorithms based on clustering according to claim 1, characterized in that, when the clustering downsampling method is the density-based spatial clustering algorithm, the first clustering condition includes: the minimum sampling value; the second clustering condition includes: the sample distance of the clusters.

4. A visualization acceleration method for particle-based algorithms based on clustering according to claim 1, characterized in that, when the clustering downsampling method is the comprehensive hierarchical clustering algorithm, the first clustering condition includes: the maximum entry tree of the leaf nodes of the clustering feature tree; the second clustering condition includes: the size of the clustering feature tree.

5. A visualization acceleration method for particle-based algorithms based on clustering according to claim 1, characterized in that, the visualization physical features include one or more of the following: the type of material, the degree of damage, the pressure.

6. A visualization acceleration method for particle-based algorithms based on clustering according to claim 5, characterized in that, before S105, it further includes the step: Obtain visualization requirements associated with the data result file, where the visualization requirements include one or more of the following: shock wave visualization, damage degree visualization; Select, from the clustering physical features in the data result file, clustering physical features that match the visualization requirements according to the visualization requirements, and mark the selected clustering physical features as visualization physical features.

7. A visualization acceleration method for a particle-based algorithm based on clustering according to claim 1, characterized in that, it further includes the steps of: S106 Obtain sampling data features from the sampling results of clustering downsampling, where the sampling data features include: the position of the cluster center, and the number represented by this cluster; S107 Calculate discard point features during the clustering downsampling process through the sampling data features, where the discard point features include: the number of discard points, and / or the distribution of discard points; S108 Determine whether the discard point features match the original data features of the data result file. If so, perform visualization calculation according to the sampling results; if not, correct the clustering input value, and perform step S105 according to the corrected clustering input value.

8. A visualization acceleration system for a particle-based algorithm based on clustering, characterized in that, it includes: A data acquisition module configured to obtain a data result file, where the data result file includes: original data for simulating the explosion damage process of an explosion object; A downsampling method matching module configured to obtain the sampling scenario features of the data result file, and find a clustering downsampling method that matches the sampling scenario features in a clustering downsampling method library; where the sampling scenario features include one or more of the following: the type of the explosion object, the time period of the explosion damage process, the cohesion characteristics of the original data, the amount of the original data, the attribute value of the original data; the clustering downsampling method library includes one or more of the following methods: k-means clustering algorithm, density-based spatial clustering algorithm, comprehensive hierarchical clustering algorithm; A first clustering condition setting module configured to select, from a first clustering condition database, a clustering setting value of a first clustering condition that matches the sampling scenario features and / or the clustering downsampling method matched in the sampling scenario features and / or the sampling method matching module, where the first clustering condition is used to adjust the number of clusters during the clustering downsampling process; A second clustering condition setting module configured to select a second clustering condition input by the user through the clustering downsampling method, and obtain a clustering input value input by the user for the second clustering condition, where the second clustering condition is used to adjust the density of clusters during the clustering downsampling process; The downsampling module is configured to use the clustering set value and the clustering input value to perform clustering downsampling on the visual physical features of the data result file through a corresponding clustering downsampling method; wherein, the second clustering condition serves as an artificial adjustment port, and the first clustering condition and the visual physical features are automatically set through a normalization program.

9. A visualization acceleration system for a particle-based algorithm based on clustering according to claim 8, wherein, when the clustering downsampling method is the k-means clustering algorithm, the first clustering condition includes: the number of initial clustering centers, and the second clustering condition includes: the number of clusters; when the clustering downsampling method is the density-based spatial clustering algorithm, the first clustering condition includes: the minimum sampling value; the second clustering condition includes: the sample distance of the cluster; when the clustering downsampling method is the comprehensive hierarchical clustering algorithm, the first clustering condition includes: the maximum entry tree of the leaf nodes of the clustering feature tree; the second clustering condition includes: the size of the clustering feature tree.

10. A visualization acceleration system for a particle-based algorithm based on clustering according to claim 8, wherein, the visual physical features include one or more of the following: the type of material, the degree of damage, and the pressure.

Images