A ship CFD cloudification simulation data management method for realizing data lightweight

Abstract

The application discloses a ship CFD cloud simulation data management method for realizing data lightweight, relates to the technical field of ship cloud simulation, and is applied to a cloud scene of ship simulation software. The method analyzes a simulation result file, stores mesh element data and node physical quantity data of a first simulation moment, and only needs to store mesh element change data and node physical quantity data at different simulation moments in subsequent numerical calculation. In addition, the node physical quantity data of each simulation moment is subjected to discretization processing, only the data of basic variables is reserved, and derivative variables are not stored. The method can greatly improve the storage data compression rate in the CFD calculation cloud scene on the basis of meeting the data display requirements of the client, greatly reduces the data storage amount on the cloud server, reduces the storage pressure of the cloud server, and improves the data transmission performance and data loading performance between the server and the client.

Description

Technical Field

[0001] This application relates to the field of ship cloud simulation technology, and in particular to a data management method for ship CFD cloud simulation that achieves data lightweighting. Background Technology

[0002] With the development of cloud technology, the cloud application of industrial CAE software has become a hot topic of widespread interest in the industry, and the cloudification of CAE software in the shipbuilding field is also gradually developing. However, the amount of ship simulation data generated by ship CAE software is enormous. Taking the cloud application of CFD for the resistance performance of a standard model DTMB5415 as an example, calculating a speed curve (including 5 speed point conditions) requires approximately 25GB of storage space. Storing the massive amount of ship simulation data on cloud servers not only occupies a large amount of physical storage space, but also easily leads to lag and crashes when clients perform real-time dynamic interaction and data visualization, seriously restricting the development of cloud applications for ship CAE software. Summary of the Invention

[0003] This application addresses the aforementioned problems and technical requirements by proposing a lightweight ship CFD cloud simulation data management method. The technical solution of this application is as follows:

[0004] A method for managing cloud-based ship CFD simulation data to achieve data lightweighting, wherein the method for managing cloud-based ship CFD simulation data executed by the cloud server includes:

[0005] The simulation result files obtained from CFD simulation using ship simulation software are parsed, and the core simulation data in the simulation result files are selected and stored. The core simulation data includes the mesh element data at the first simulation time, the nodal physical quantity data at the first simulation time, and the mesh element change data and nodal physical quantity data at each other simulation time. Among them, the mesh element change data at each simulation time is the change data of the mesh element data at the simulation time relative to the mesh element data at the previous simulation time, and the nodal physical quantity data at each simulation time is the data of the basic variables selected from the original result data at each simulation time.

[0006] When a data acquisition request containing target display data is received from a client, the target display data for each simulation moment is obtained by converting the stored core simulation data, and the target display data for each simulation moment is sent to the client for display.

[0007] The beneficial technical effects of this application are:

[0008] This application discloses a data management method for cloud-based ship CFD simulation that achieves data lightweighting. Applied to cloud-based scenarios for ship simulation software, this method parses the simulation result file and stores the mesh element data and node physical quantity data for the first simulation moment. In subsequent simulation moments, only the mesh element change data and node physical quantity data for each simulation moment need to be stored, eliminating the need to repeatedly store the mesh topology. Furthermore, the node physical quantity data for each simulation moment is discretized, retaining only the basic variables and not storing derived variables. Derived variables are calculated only when the client needs to display the data. This method significantly improves the data compression rate in cloud-based CFD computing scenarios while meeting the client's data display requirements, greatly reducing the amount of data stored on the cloud server. This not only reduces the storage pressure on the cloud server but also improves the data transmission performance between the server and the client, optimizing the client's data loading performance and ensuring smooth loading and display of simulation results.

[0009] This method further filters redundant data in the core simulation data, further reducing the data size. The dynamic node averaging variable storage method developed in this application can dynamically generate vertices, saving storage space. Considering the different numerical accuracy requirements of clients, this method also employs a precision-controllable lossy data compression algorithm based on numerical accuracy requirements to actively reduce the precision of the large amount of high-precision data generated during numerical calculations. This mainly includes reducing the calculation frame rate and extracting the mesh density, thereby further achieving significant compression of simulation result data and improving data transmission and loading performance. Through the combination of multiple compression strategies, this method significantly reduces the storage space of ship CAE cloud data, ensures smooth loading and display of transient results, and provides a technical foundation for ship CAE cloudification. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the method flow of a ship CFD cloud simulation data management method according to an embodiment of this application.

[0011] Figure 2 This is a schematic diagram of the average storage of dynamic nodes.

[0012] Figure 3 This is a schematic diagram of the topological structure of the mesh element data extracted at the first simulation moment in an instance.

[0013] Figure 4 yes Figure 3 This example illustrates the transformation of isosurface nodes based on the same mesh element data at different simulation times. Detailed Implementation

[0014] The specific embodiments of this application will be further described below with reference to the accompanying drawings.

[0015] This application discloses a method for managing cloud-based ship CFD simulation data to achieve data lightweighting. This method is applied to cloud-based scenarios for ship simulation software, which is deployed and runs on a cloud server. The data generated during the simulation is stored on the cloud server and then transmitted to the client for display. The ship CFD cloud-based simulation data management method of this application is executed by the cloud server and includes the following:

[0016] Ship simulation software outputs simulation result files during CFD simulations. These files include mesh element data for each simulation moment and the original result data for each mesh node, sometimes along with other simulation data, resulting in a large file size. In practice, for transient simulation results of ship performance, different frames are typically stored in different simulation result files. After data conversion, these files exist in different cloudified data sets, requiring cloudification and merging. This application improves data lightweighting through the following solutions:

[0017] 1. Filtering in a general grid format

[0018] Traditionally, simulation result files output by ship simulation software are directly stored. However, this application does not store them directly. Instead, it first parses the simulation result files obtained by the ship simulation software, filters out the core simulation data in the simulation result files, and then stores only the core simulation data. The remaining data in the simulation result files are not stored. The remaining data includes data that can be converted from the core simulation data and redundant data that has no impact on the simulation results within the error range.

[0019] The core simulation data extracted in this application includes mesh element data and nodal physical quantity data at the first simulation time step, as well as mesh element change data and nodal physical quantity data at each of the other simulation time steps. Specifically, the mesh element change data at each simulation time step represents the change in mesh element data at the current simulation time step relative to the mesh element data at the previous simulation time step.

[0020] In ship simulation software, the simulation result files output include mesh element data for each simulation moment. Therefore, directly storing the simulation result files would be equivalent to storing mesh element data for all simulation moments. However, CFD calculations use Eulerian meshes. Although the mesh undergoes compression and stretching during numerical calculations due to ship motion, the mesh topology remains essentially unchanged; that is, the mesh topology is consistent across different simulation moments. Based on this idea, this application does not store mesh element data separately for each simulation moment. Instead, it stores only the mesh element data for the first simulation moment as a general mesh topology. For subsequent simulation moments starting from the second moment, instead of repeatedly storing mesh element data, it stores the changing mesh element data—that is, the differentiated parts. Thus, only one set of mesh topology needs to be stored, significantly reducing the data size.

[0021] 2. Removal of derived variables

[0022] In addition, each simulation moment outputs raw result data for each grid node. This raw result data contains numerous variables, and some of these variables are correlated. Therefore, this application categorizes the variables in the raw result data for each grid node into two types: basic variables and derived variables. Derived variables are those that can be derived from basic variables. Since derived variables can be derived from basic variables, during storage, the basic variable data is selected from the raw result data of each grid node at each simulation moment and stored as the node physical quantity data for that simulation moment. The derived variable data for that simulation moment is not stored. Based on the actual situation of CFD calculations, the basic variables include grid node number, grid node displacement, grid node velocity, and grid node pressure; the derived variables include streamlines and pathlines.

[0023] 3. Removal of redundant information in the fluid domain

[0024] Furthermore, the simulation results output by ship simulation software are generally obtained by simulating the entire fluid domain. This fluid domain typically covers the entire area where the ship is located and a portion of the fluid it inhabits. Some meshes within the fluid domain will affect the ship's simulation results, while others will not. Therefore, based on the mesh topology indicated by the mesh element data at the first simulation moment, meshes in redundant regions outside the target region of the flow field are deleted, while the core simulation data within the target region of the flow field is retained and stored. The target region of the flow field is the area that affects the ship's simulation performance, while the redundant region is the area that does not affect the ship's simulation performance. Based on the actual simulation, the redundant region includes the area above the ship's deck, the area along the ship's bow to the front of the fluid domain, and the ship's interior. This operation further filters the core simulation data before storage, further reducing the amount of data that needs to be stored.

[0025] Therefore, this application, through the lightweight data design of the aforementioned multiple stages, can significantly reduce the data storage pressure on the cloud server. The data stored on the cloud server is used to transmit data to the client for display. When the client needs to perform real-time dynamic interaction and data visualization, the client sends a data acquisition request containing the target display data to the cloud server. The target display data is the ship CFD cloud simulation data that the client needs to display.

[0026] When the cloud server receives a data acquisition request containing target display data from the client, it converts the stored core simulation data to obtain the target display data for each simulation moment, and sends the target display data for each simulation moment to the client for display.

[0027] When the target display data required by the client only includes the basic variables from the original results data at each simulation moment, the cloud server transforms the stored core simulation data to obtain the basic variable data for each simulation moment. This includes: determining the information of the grid nodes at the first simulation moment based on the grid element data, and combining this with the node physical quantities at the first simulation moment to obtain the basic variable data for each grid node at the first simulation moment. For any simulation moment starting from the second simulation moment, the grid element data for the current simulation moment is obtained by combining the grid element change data at the current simulation moment with the grid element data from the previous simulation moment; the information of the grid nodes at the current simulation moment is determined based on the grid element data at the current simulation moment; and the basic variable data for each grid node at the current simulation moment is updated based on the information of the grid nodes at the current simulation moment and the node physical quantity data at the current simulation moment.

[0028] As mentioned above, the mesh topology remains unchanged at different simulation moments, but the actual mesh still changes as the simulation progresses, with changes occurring in mesh positions and other information. Since the storage process stores mesh element changes starting from the second simulation moment, from the second simulation moment onwards, it is necessary to generate the current simulation moment's mesh based on the mesh from the previous simulation moment, and then assign the node physical quantity data from the current simulation moment to the current simulation moment's mesh. This embodiment provides the following dynamic node averaging variable storage method to achieve the above process.

[0029] 4. Variable storage for dynamic node averaging

[0030] (1) The grid element data and grid element change data at each simulation time are stored at the grid center point, such as Figure 2 E-type data in the data structure typically includes grid cell identifiers, node lists, grid element types, and material properties, among other things. Figure 2 Taking mesh elements, including fluid elements, solid elements, shell elements, membrane elements, beam elements, and rod elements, as an example.

[0031] By using the nodal averaging method (interpolation function and shape function) to utilize the grid element data at each grid center point, information about each grid node at the current simulation moment can be dynamically interpolated. This step can deduce the vertices around the grid from the grid center point. For example, the nodes of the water surface around the hull (where the boundary in contact with the hull changes) can be deduced from the nodes of the hull shell (which does not deform but undergoes displacement).

[0032] (2) After obtaining the grid center point and grid vertices, the data of the basic variables of each grid node in the current simulation time are dynamically interpolated using the dynamic triangulation algorithm based on the node physical quantity data at the current simulation time, thereby updating the data of each grid node in the current simulation time:

[0033] (a) Construct a super triangle and initialize the triangular mesh. The super triangle contains all the mesh nodes at the current simulation time.

[0034] (b) For each mesh node p at the current simulation time, calculate the circumcircle of each triangle in the current triangulation, determine all triangles whose circumcircles contain mesh node p, determine the polygonal holes formed by the triangles whose circumcircles contain mesh node p, delete the triangles whose circumcircles contain mesh node p from the triangulation, connect mesh node p to each vertex of the polygonal hole to form a new triangle and add it to the triangulation, thus completing the update of the triangulation.

[0035] (c) After updating the triangulation based on each grid node p, delete all triangles in the triangulation that are related to the vertices of the super triangle, and dynamically generate the node data sequence based on the final updated triangulation, and then dynamically generate the data of each grid node p.

[0036] For example, in a transient CFD simulation of a ship traveling in water at different speeds, the isosurface waveform of the water surface changes at each moment. The topological structure of the mesh element data of the isosurface extracted at the first simulation moment is as follows: Figure 3 As shown. Based on this unified mesh element data, by using the mesh element change data between adjacent simulation times, the isosurface of the next simulation time T1 can be formed from simulation time T0, thus unifying the isosurface mesh at different times. Please refer to [reference needed]. Figure 4 The diagram illustrates how this reduces the space required for data storage.

[0037] In addition to the basic variable data, the target display data required by the client may also have other display requirements. Therefore, the cloud server needs to perform data transformation based on the basic variable data to meet the client's display needs.

[0038] 5. Calculation of derived variables

[0039] When the target display data that the client needs to display also includes data of derived variables, the data of the basic variables for each simulation moment obtained by the cloud server based on the core simulation data stored also includes: data of derived variables obtained by converting the data of the basic variables of each grid node for each simulation moment.

[0040] As mentioned above, in CFD computation scenarios, the derived variables mainly include streamlines and pathlines, and the calculation methods are as follows:

[0041] (1) Calculation of streamlines

[0042] Initialize the position of the first discrete point x0 of the streamline, and then determine the position of any i-th discrete point x0. i according to The position x of the (i+1)th discrete point is calculated. i+1 until the length of the streamline reaches the target length or until the (i+1)th discrete point position x i+1 Upon reaching the predetermined end region of the streamline, the positions x of each discrete point are obtained. i The streamlines formed.

[0043] in, The integer parameter i starts at 0. Any V(x) represents the velocity of the mesh node at position x, and Δt represents the time interval between two adjacent simulation moments.

[0044] (2) Calculation of trace

[0045] Initialize the position y0 of the first discrete point of the trace and the starting time t0, and then determine the position y0 of any j-th discrete point. j according to The position y of the (j+1)th discrete point is calculated. i+1 until the length of the trace reaches the target length of the trace or until the position of the (j+1)th discrete point y. i+1 Upon reaching the predetermined end region of the trace, the positions y of each discrete point are obtained. j The traces formed.

[0046] in, The integer parameter j starts at 0. Any V(y, t) represents the velocity of a grid node at position y at time t.

[0047] 6. Lossy data compression algorithm with controllable accuracy

[0048] Furthermore, to ensure computational accuracy, the simulation result files output by ship simulation software typically contain a large number of meshes and data attached to those meshes. However, in post-processing and client-side visualization, such a large number of high-precision meshes are often unnecessary, and clients may have different display requirements in different scenarios. For example, the requirements for numerical accuracy differ depending on whether the client needs to display at 35 frames per second or at 20 frames per second; the client has lower requirements for numerical accuracy when displaying at 20 frames per second. Therefore, in one embodiment, the data acquisition request received by the cloud server from the client also includes a numerical accuracy requirement, indicating the client's precision requirements. The cloud server then performs lossy compression on the target display data for each simulation moment obtained from the data acquisition request based on the numerical accuracy requirement before sending it to the client. This operation can achieve significant compression while meeting the numerical accuracy requirements, reducing storage pressure.

[0049] The cloud server transmits the compressed target display data to the client, where it is decompressed and rendered. The compressed target display data stored by the cloud server is used for long-term storage and rendering visualization.

[0050] This application, through a combination of the above strategies, can significantly improve the data compression rate in cloud-based CFD computing scenarios while ensuring that the data display needs of the client are met. This greatly reduces the amount of data stored on the cloud server, not only reducing the storage pressure on the cloud server but also improving the data transmission performance between the server and the client, and optimizing the client's data loading performance, ensuring smooth loading and display of simulation results. In one example, compared with the examples in the background technology, the method of this application can reduce the data storage size of the simulation result file generated from calculating a speed curve (containing 5 speed point conditions) for the standard model DTMB5415 from 25GB to 12GB while meeting the client's data display needs.

[0051] The above descriptions are merely preferred embodiments of this application, and this application is not limited to the above embodiments. It is understood that other improvements and variations that can be directly derived or conceived by those skilled in the art without departing from the spirit and concept of this application should be considered to be included within the protection scope of this application.

Claims

1. A method for managing cloud-based ship CFD simulation data to achieve lightweight data, applied to a cloud server, characterized in that: The ship CFD cloud simulation data management method includes: The simulation result file obtained from CFD simulation using ship simulation software is parsed, and the core simulation data in the simulation result file is selected and stored. The core simulation data includes the mesh element data and node physical quantity data at the first simulation time, as well as the mesh element change data and node physical quantity data at each other simulation time. The mesh element change data at each simulation time is the change data of the mesh element data at that simulation time relative to the mesh element data at the previous simulation time, and the node physical quantity data at each simulation time is the data of the basic variables selected from the original result data at that simulation time. When a data acquisition request containing target display data is received from a client, the target display data for each simulation moment is obtained by converting the stored core simulation data, and the target display data for each simulation moment is sent to the client for display. The target display data is the ship CFD cloud simulation data that the client needs to display. The process of converting the stored core simulation data to obtain the target display data for each simulation moment includes: determining the information of the grid nodes for the first simulation moment based on the grid element data of the first simulation moment, and converting the data of the basic variables of each grid node for the first simulation moment by combining the node physical quantities of the first simulation moment; for any simulation moment starting from the second simulation moment, converting the grid element change data of the current simulation moment by combining the grid element data of the previous simulation moment to obtain the grid element data of the simulation moment, determining the information of the grid nodes for the current simulation moment based on the grid element data of the current simulation moment, and updating the data of the basic variables of each grid node for the current simulation moment based on the information of the grid nodes for the current simulation moment and the node physical quantity data of the current simulation moment.

2. The ship CFD cloud simulation data management method according to claim 1, characterized in that, The ship CFD cloud simulation data management method also includes: Based on the mesh topology indicated by the mesh element data at the first simulation moment, the meshes in redundant regions outside the target region of the flow field are deleted, while the core simulation data within the target region of the flow field are stored. The target region of the flow field is the region that affects the ship simulation performance.

3. The ship CFD cloud simulation data management method according to claim 1, characterized in that, Send the target display data for each simulation moment to the client, including: Based on the numerical accuracy requirements indicated by the data acquisition request, the target display data for each simulation moment obtained by conversion is subjected to lossy compression processing and then sent to the client.

4. The ship CFD cloud simulation data management method according to claim 1, characterized in that, The original result data for each simulation time step in the simulation result file includes not only the data of the basic variables, but also the data of the derived variables, which are variables transformed from the basic variables; The process of converting the stored core simulation data to obtain the target display data for each simulation moment also includes: converting the data of the basic variables of each grid node at each simulation moment to obtain the data of the derived variables.

5. The ship CFD cloud simulation data management method according to claim 4, characterized in that, The basic variables include grid node number, grid node displacement, grid node velocity, and grid node pressure, while the derived variables include streamlines and pathlines.

6. The ship CFD cloud simulation data management method according to claim 5, characterized in that, The resulting streamlines and traces include: Position of the first discrete point of the initialized streamline And according to any first Location of discrete points according to Calculate the first Location of discrete points Until the length of the streamline reaches the target length of the streamline or until the first... Location of discrete points Upon reaching the predetermined end region of the streamline, the positions of each discrete point are obtained. The streamlines formed; among them, Integer parameters The initial value is 0; any Indicates position The speed of the grid nodes at that location This represents the time interval between two adjacent simulation moments; Initialize the position of the first discrete point of the trace and start time And according to any first Location of discrete points according to Calculate the first Location of discrete points Until the length of the trace reaches the target length or until the first... Location of discrete points Upon reaching the predetermined end region of the trace, the positions of each discrete point are obtained. The traces formed; among them, Integer parameters The initial value is 0; any Indicates position At the moment The speed of the grid nodes at that time.

7. The ship CFD cloud simulation data management method according to claim 1, characterized in that, The grid element data and grid element change data at each simulation moment are stored at the grid center point; The information of the mesh nodes at the current simulation moment, determined based on the mesh element data at the current simulation moment, includes: By using the node averaging method, the information of each grid node at the current simulation moment is obtained by dynamically interpolating the grid element data at the center point of each grid.

8. The ship CFD cloud simulation data management method according to claim 1, characterized in that, Update the data of the basic variables of each grid node at the current simulation moment, including: Based on the node physical quantity data at the current simulation moment, the basic variable data of each grid node at the current simulation moment is obtained by dynamic interpolation using a dynamic triangulation algorithm.

9. The ship CFD cloud simulation data management method according to claim 8, characterized in that, The data of the basic variables of each grid node at the current simulation moment obtained by dynamic interpolation using the dynamic triangulation algorithm include: Construct a super triangle and initialize the triangular mesh, wherein the super triangle contains all the mesh nodes in the current simulation time; For each mesh node at the current simulation moment, calculate the circumcircle of each triangle in the current triangulation, determine all triangles whose circumcircles contain the mesh node, determine the polygonal holes formed by the triangles whose circumcircles contain the mesh node, delete the triangles whose circumcircles contain the mesh node from the triangulation, connect the mesh node to each vertex of the polygonal hole to form a new triangle and add it to the triangulation, thus completing the update of the triangulation. After updating the triangulation based on each grid node, all triangles related to the vertices of the super triangle are deleted from the triangulation, and the basic variable data of each grid node is dynamically generated based on the final updated triangulation.

Images