A method for multiplexing flow field data in fluid mechanics

Abstract

The present invention discloses a method for reusing fluid mechanics flow field data, which belongs to the technical field of computational fluid dynamics and aircraft design. The method quickly establishes the correspondence between new and old grid micro-elements based on a balanced multi-dimensional binary tree fast retrieval algorithm, and interpolates the flow solution physical quantities on the old grid, such as the density, velocity and pressure of the flow medium, to the new grid in parallel according to the correspondence, so as to realize the rapid establishment of the new grid space physical field. The present invention can be widely adapted to the parallel flow field reuse of various similar working conditions, can efficiently reuse the existing spatial data, and save a lot of computing resources.

Description

Technical Field

[0001] The invention belongs to the technical field of computational fluid dynamics and aircraft design, and in particular relates to a method for multiplexing fluid dynamics flow field data. Background Art

[0002] The performance requirements of computational fluid dynamics (CFD) applications are constantly evolving. With the complexity of the shape of numerical simulation objects and the need for refined flow simulation, the number of grid cells in numerical simulation can reach tens of millions, hundreds of millions, or even tens of billions. The calculation time for a single condition is long, which leads to the demand for huge computing power in CFD, and multi-state concurrency or iterative concurrency mode is often used. In the numerical simulation process of multiple concurrent computing conditions, flow field reuse is a very important computing acceleration technology. Flow field reuse means that when the computing conditions are "similar" (such as flight conditions, physical models, aircraft shape / grids, etc., the change range of computing conditions is not large), the flow field data of existing / completed conditions is used as the initial flow field of the new condition for numerical simulation. Due to the "similarity" of the two, the convergence speed of the numerical simulation of the flow field of the new condition can be accelerated to a certain extent, thereby improving the computing efficiency and reducing the consumption of computing resources.

[0003] Conventional flow field reuse methods mainly focus on the “similarity” of flight conditions: when flight conditions change to a certain extent, the new condition directly adopts the flow field of the old condition as the initial flow field, thus achieving effective flow field reuse.

[0004] Conventional flow field reuse methods rely on the one-to-one correspondence between the mesh elements and their physical variables in the new and old working conditions. This method may not be directly applicable when the aircraft shape changes or the mesh changes. For example, when the aircraft shape changes (such as a large rotation of the rudder), the one-to-one correspondence between the mesh elements near the rudder may be destroyed, and the traditional method may not be able to achieve reuse; when the aircraft mesh type changes (for example, the old working condition is an unstructured mesh, and the new working condition is a structured mesh), the traditional method may not be able to achieve reuse. Summary of the invention

[0005] The purpose of the present invention is to provide a method for reusing fluid mechanics flow field data to overcome the defects of the prior art. The corresponding relationship between the new and old grid elements is quickly established based on a balanced multidimensional binary tree fast retrieval algorithm. According to the corresponding relationship, the flow solution physical quantities on the old grid are interpolated to the new grid in parallel, thereby realizing the rapid establishment of the new grid space physical field.

[0006] The object of the present invention is achieved through the following technical solutions:

[0007] A method for multiplexing fluid mechanics flow field data, the method comprising:

[0008] Obtain the numerical simulation results of the aircraft flow field under the original working condition, store the original calculated old grid data and old dimensionless flow field data, and import the flight conditions under the new working condition and the parameters of the new grid;

[0009] Traverse the partition blocks of the new grid and the old grid, build the minimum bounding box according to the spatial position of the new grid and the old grid, and determine whether the new grid and the old grid have an intersection relationship in space;

[0010] If there is an intersection relationship, a balanced binary tree retrieval structure of the old grid is generated in a recursive mode, each grid element of the new grid is traversed, and the index of the old grid element closest to the new grid element is recursively searched in the balanced binary tree retrieval structure;

[0011] According to the index and the old dimensionless flow field data, the estimated dimensionless flow field value of the grid micro-element in the new grid is calculated;

[0012] The numerical simulation of the new working condition is completed with the estimated dimensionless flow field value, and the aerodynamic data of the aircraft under the new working condition is obtained.

[0013] Furthermore, the original calculated old grid data includes at least the dimensionless position coordinates of each grid element and the minimum bounding box range of the grid block, and also includes one or more of the grid type, grid partitions, topological relationships between partitions, sequence numbers of each element, and connection relationships between each element.

[0014] Furthermore, the old dimensionless flow field data includes one or more of gas pressure, temperature and density.

[0015] Furthermore, the parameters of the new grid include at least the grid type, grid partition, grid block number, dimensionless position coordinates of the grid element and the minimum bounding box range of the grid block, and also include one or more of the topological relationship between the grid partitions, the sequence number of each element, the type of each element and the connection relationship between each element.

[0016] Furthermore, the estimated dimensionless flow field values ​​of the grid micro-element in the new grid are calculated, including:

[0017] The spatial dimensionless flow field data corresponding to the index position is read from the dimensionless flow field data, and assigned to the corresponding spatial point position of the new grid element. The assignment method includes direct assignment or distance-based spatial interpolation.

[0018] Furthermore, generating a balanced binary tree retrieval structure of the old grid in a recursive mode specifically includes:

[0019] All elements in the old grid are divided into multiple finite point subsets for storage, and each finite point subset forms a one-to-one mapping relationship with the relevant nodes on the balanced binary tree; among them, the finite point subset contained in each non-terminal node is the union of the finite point subsets contained in its left child and right child nodes, and the finite point subsets contained in the same-level brother nodes do not intersect with each other.

[0020] Furthermore, recursively searching for the index of the old grid element closest to the new grid element in the balanced binary tree search structure specifically includes:

[0021] For each microelement of the new grid, the distance between microelement of the old grid is recursively calculated in the balanced binary tree retrieval structure, and the nearest microelement of the old grid is located to establish the index of the nearest microelement of the old grid.

[0022] Furthermore, when the method calculates the estimated dimensionless flow field value of the grid element in the new grid based on the index and the old dimensionless flow field data, the old dimensionless flow field data is read using parallel reading and writing technology. When reading in data, only the old grid data and dimensionless flow field data that have an intersection relationship with the grid stored in the local process are read in.

[0023] Furthermore, judging whether the new grid and the old grid have an intersection relationship in space specifically includes:

[0024] The minimum bounding box includes the minimum and maximum data of all the infinitesimal elements of the grid block along the coordinate direction in the Cartesian coordinate axis. By comparing the minimum bounding boxes of the new and old grid blocks, the grid blocks that do not touch or intersect in the geometric space are excluded.

[0025] Furthermore, the method uses the HDF5 standard data format to store grid data and dimensionless flow field data.

[0026] The beneficial effects of the present invention are:

[0027] The present invention quickly establishes the correspondence between the new and old grid micro-elements based on the balanced multi-dimensional binary tree fast retrieval algorithm, and interpolates the flow solution physical quantities on the old grid, such as the density, velocity and pressure of the flow medium, to the new grid according to the correspondence, so as to realize the rapid establishment of the new grid space physical field. It can be widely adapted to the flow field reuse of various working conditions with similar flight conditions, aircraft shapes, computational grid partitions / topologies / types, etc., and can efficiently reuse existing spatial data, improve computing efficiency, and save a lot of computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flowchart of a method for multiplexing fluid mechanics flow field data according to an embodiment of the present invention;

[0029] Figure 2 is a schematic diagram of an old working condition grid of an embodiment of the present invention;

[0030] Figure 3 is a schematic diagram of a new working condition grid of an embodiment of the present invention;

[0031] Figure 4 This is a Mach number cloud diagram of a flow field in a sparse grid numerical simulation under an old working condition according to an embodiment of the present invention;

[0032] Figure 5 The Mach number cloud diagram of the initial flow field under the new working condition is obtained in parallel by the eighth core of the embodiment of the present invention. DETAILED DESCRIPTION

[0033] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0034] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without making any creative work shall fall within the scope of protection of the present invention.

[0035] Conventional flow field reuse methods rely on the one-to-one correspondence between the mesh elements and their physical variables in the new and old working conditions. This method may not be directly applicable when the aircraft shape changes or the mesh changes. For example, when the aircraft shape changes (such as a large rotation of the rudder), the one-to-one correspondence between the mesh elements near the rudder may be destroyed, and the traditional method may not be able to achieve reuse; when the aircraft mesh type changes (for example, the old working condition is an unstructured mesh, and the new working condition is a structured mesh), the traditional method may not be able to achieve reuse.

[0036] In order to solve the above technical problems, the following embodiments of a method for multiplexing fluid mechanics flow field data are proposed in the present invention.

[0037] Example 1

[0038] Reference Figure 1 ,like Figure 1 The figure is a flowchart of a method for multiplexing fluid mechanics flow field data in this embodiment, and the method specifically includes the following steps:

[0039] Step 1: Obtain the numerical simulation result data of the aircraft flow field under the original working condition, store the parallel partitioned grid and the corresponding dimensionless flow field of the original calculation; and import the flight conditions and grid of the new working condition.

[0040] This step requires obtaining the original aircraft flow field simulation result data and storing it in the required data format: First, obtain the numerical simulation data of the aircraft's original working condition flow field. The old grid can be of any type, such as different grid partitions / different grid distributions / different grid densities / different grid types. Perform dimensionless processing on the original calculation data to obtain various parameters on each grid element of the flow field, including grid parameters and flow field parameters, and store them.

[0041] The parameters of the above-mentioned old grid may include, but are not limited to, the dimensionless position coordinates of each grid element (including X-axis coordinates, Y-axis coordinates, Z-axis coordinates, etc.), the minimum bounding box range of the grid block, etc., and may also include characteristic parameters such as grid type, grid partition, topological relationship between partitions, sequence number of each element, and connection relationship between each element. Among them, the dimensionless parameters of the grid can be dimensionless with reference to the characteristic dimensions of the aircraft. The grid storage data format adopts the HDF5 standard data format, the core of which is the group structure and data set. It is a cross-platform parallel data storage format that facilitates quick and convenient finding of required data from large data files. The use of this data format enables this patent to achieve large-scale data reuse with less memory requirements.

[0042] The dimensionless parameters of the flow field physical quantities mentioned above may include but are not limited to dimensionless gas pressure, temperature, density, etc. The storage format of the flow field physical quantity parameters is consistent with the grid parameter storage format, and maintains a one-to-one correspondence with the grid element index, which facilitates the subsequent rapid search for required data.

[0043] The parameters of the above new grid may include but are not limited to grid type, grid partition, grid block number, dimensionless position coordinates of grid micro-element (including X-axis coordinate, Y-axis coordinate, Z-axis coordinate, etc.), minimum bounding box range of grid block, etc., as well as characteristic parameters such as topological relationship between grid partitions, sequence number of each micro-element, type of each micro-element (boundary micro-element, internal micro-element, extended micro-element, virtual micro-element, etc.) and connection relationship between each micro-element. In order to meet the needs of large-scale parallel data use, the new grid is allocated to multiple parallel processes according to the partition requirements, and each process stores the grid parameters in the corresponding numbered new grid block.

[0044] Step 2: Traverse the new and old grid partition blocks, and build the minimum bounding box of the new and old grids according to the spatial positions of the new and old grid blocks to quickly determine whether the new and old grids overlap or touch each other in space.

[0045] This step uses the minimum bounding box of the old and new grid blocks to quickly determine whether there is an intersection relationship. The old grid has stored the minimum bounding box information in step 1, which can be quickly read from the stored old grid data. The new grid obtains the bounding box information by quickly calculating the grid micro-element traversal. The minimum bounding box information parameters are the minimum and maximum values ​​of all micro-elements of the grid block along the coordinate direction in the Cartesian coordinate axis. By comparing the minimum bounding box information of the old and new grid blocks, grid blocks that do not touch or intersect in the geometric space can be quickly excluded, reducing the amount of subsequent data search.

[0046] Step 3: If the old grid and the new grid intersect in space, generate a balanced binary tree retrieval structure of the old grid in a recursive mode; traverse each grid element of the new grid, and recursively search for the index of the old grid element closest to the new grid element in the balanced binary tree retrieval structure.

[0047] The balanced binary tree-based retrieval method of the old grid established in this step is a key technology for the embodiment of the present invention to realize fast flow field data interpolation and reuse. The balanced binary tree retrieval structure is a tree data structure that stores instance points in a multidimensional space for fast retrieval. Constructing a balanced binary tree retrieval structure is equivalent to continuously dividing the multidimensional space with hyperplanes perpendicular to the coordinate axis to form a series of multidimensional hyperrectangular regions, each of which corresponds to a multidimensional hyperrectangular region. Using a balanced binary tree retrieval structure can save the search for most data points, thereby reducing the amount of search calculations.

[0048] For each micro-element of the new grid, the distance from the micro-element of the old grid to it is recursively calculated in the established balanced binary tree, and the nearest micro-element of the old grid is located, and the index of the nearest micro-element of the old grid is established.

[0049] In this embodiment, the specific method of generating a balanced binary tree search structure in a recursive mode is as follows:

[0050] First, suppose that The given point set of point elements is , define a A vector container ( ), sequential storage Points The address value is recorded as , where the class describing the point element information contains two basic data: ① The spatial coordinates of the point element, in the form of a double-precision floating-point array The identifier, its length and spatial dimension Same;②Point elements in The initial sequence number in the signed integer variable Identification; at a given point set In the example, any two different point elements have different value.

[0051] Vector All elements in are divided into multiple finite point subsets for storage, and each finite point subset forms a one-to-one mapping relationship with the relevant nodes on the balanced binary tree; among them, the finite point subset contained in each non-terminal node is the union of the finite point subsets contained in its left child and right child nodes, and the finite point subsets contained in the same-level brother nodes do not intersect with each other.

[0052] by As a binary tree node type name, The internal member variables of include: ① The finite point subset elements contained in the current node are Starting subscript in vector ;②The finite point subset elements contained in the current node are End subscript in vector ;③The finite point subset contained in the current node is The infimum array in the directions ; ④ The finite point subset contained in the current node is The supremum array in the directions ; ⑤The left child pointer of the current node , whose initialization value is a null pointer ;⑥The right child pointer of the current node , whose initialization value is a null pointer .

[0053] Set a finite point subset length threshold for a balanced binary tree and use a signed integer member variable Store; if the current node contains a finite number of point subset elements Satisfy the conditions , then the current node is identified as a non-terminal node and needs to continue recursive decomposition; otherwise, the current node is identified as a terminal node.

[0054] Declare a binary tree recursive generator function with no return type and its parameter list. The recursive function is named , its input parameters are: ① subtree root node pointer ;② The node contains the point subset elements in The starting subscript in ; ③ The node contains the point subset elements in The end subscript in ; then the recursive function is in the form of .

[0055] The global root node pointer of the balanced binary tree Constructing instantiated objects: In the binary tree constructor, the global root node pointer The initial value is the empty address , Passing by reference to a recursive function The first parameter of At the same time, let the formal parameter and The values ​​are 0 and ;in, For vector The total length of

[0056] Among them, dynamic memory is opened to achieve the current Object instantiation, and through the node type The assignment operation inside the constructor uses Private member variables of an object , The values ​​of Parameters , Be consistent.

[0057] Along The current coordinate axis directions are calculated separately The point elements contained in the object The variance of the coordinate components of the Record the direction number corresponding to the maximum variance; among them, is an integer and .

[0058] Along the direction of maximum variance, according to the coordinate components of the point elements The rules from small to large are currently The point element address value contained in the object .

[0059] Calculate the root node of the current subtree The upper and lower bounds of the bounding box of the space. coordinate axis directions, respectively calculate the current The point elements contained in the object The coordinate components The infimum of Supremum ;in, is an integer and .

[0060] Instantiation of subtrees at all levels of a balanced binary tree. It should be noted that before executing the instantiation step of subtrees at all levels of a balanced binary tree, if the condition If true, then the recursive function Return immediately, otherwise continue to execute the instantiation steps of subtrees at all levels of the balanced binary tree and then return.

[0061] Step 4: Based on the index established in step 3, the dimensionless flow field data of the old flow field is read from the old flow field file using parallel reading and writing technology, and the estimated dimensionless flow field value of the grid element in the new grid is calculated by interpolation.

[0062] This step quickly obtains the dimensionless flow field data of the spatial position from the corresponding index in the old grid. According to the index established in step 3, the spatial dimensionless flow field data of the index position is read from the old flow field data established in step 1 using parallel reading and writing technology, and assigned to the corresponding spatial point position of the new grid micro-element. The assignment method can be direct assignment or distance-based spatial interpolation, so that the parameter value of the estimated flow field on the new grid micro-element of the new working condition can be obtained.

[0063] The above-mentioned parallel reading and writing technology allows only the old grid information and physical field information that intersect with the grid stored in the local process to be read in when data is read in. On the one hand, it can significantly reduce the reading time, and on the other hand, it can reduce the memory usage requirements. It is suitable for reading large-scale grid information and physical field data.

[0064] In the case where the new grid element indexed by the old grid element cannot be established in step 3, the dimensionless physical quantity parameters of the spatial position of the element are given according to the new flight conditions given in step 1.

[0065] Step 5: Based on the estimated dimensionless flow field value of the new grid, complete the numerical simulation of the new working condition, and then obtain various aerodynamic data of the aircraft under the new working condition.

[0066] This embodiment quickly establishes the correspondence between the new and old grid micro-elements based on the balanced multi-dimensional binary tree fast retrieval algorithm, and interpolates the flow solution physical quantities on the old grid, such as the density, velocity and pressure of the flow medium, to the new grid according to the correspondence, so as to realize the rapid establishment of the new grid space physical field. It can be widely adapted to the reuse of flow fields of various similar working conditions, can efficiently reuse existing spatial data, improve computing efficiency, and save a lot of computing resources.

[0067] Example 2

[0068] This embodiment takes the HyFLOW software in the National Numerical Wind Tunnel Project as an example to further describe the technical solution of the above embodiment. Here, a variable bottom shape aircraft is taken as an example to illustrate the application effect of the fluid mechanics flow field data reuse method provided by the above embodiment.

[0069] Reference Figure 2 and Figure 3 ,like Figure 2The figure shows a schematic diagram of the old working condition grid of this embodiment. Figure 3 Shown is a schematic diagram of the new working condition grid of this embodiment.

[0070] From the Z=0 section grid, it can be seen that the shape of the aircraft corresponding to the new and old grids has changed, and the grid topology and distribution have also changed. Since there is no one-to-one correspondence between the micro-elements of the new and old grids, the conventional flow field reuse method cannot be used effectively. However, the fluid mechanics flow field data reuse method provided by the above embodiment can achieve better flow field reuse. Figure 4 and Figure 5 ,like Figure 4 The figure shows the Mach number cloud diagram of the flow field of the old working condition sparse grid numerical simulation of the embodiment of the present invention. Figure 5 The figure shows the Mach number cloud diagram of the initial flow field under the new working condition obtained by parallel operation of the 8th core of the embodiment of the present invention.

[0071] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for multiplexing fluid mechanics flow field data, characterized in that: The method comprises: Obtain the numerical simulation results of the aircraft flow field under the original working condition, store the original calculated old grid data and old dimensionless flow field data, and import the flight conditions under the new working condition and the parameters of the new grid; Traverse the partition blocks of the new grid and the old grid, build the minimum bounding box according to the spatial position of the new grid and the old grid, and determine whether the new grid and the old grid have an intersection relationship in space; If there is an intersection relationship, a balanced binary tree retrieval structure of the old grid is generated in a recursive mode, each grid element of the new grid is traversed, and the index of the old grid element closest to the new grid element is recursively searched in the balanced binary tree retrieval structure; The balanced binary tree retrieval structure of the old grid generated in recursive mode specifically includes: All elements in the old grid are divided into multiple finite point subsets for storage, and each finite point subset forms a one-to-one mapping relationship with the relevant nodes on the balanced binary tree; the finite point subset contained in each non-terminal node is the union of the finite point subsets contained in its left child and right child nodes, and the finite point subsets contained in the sibling nodes at the same level do not intersect with each other; The index of the old grid element closest to the new grid element is recursively searched in the balanced binary tree search structure, specifically including: For each microelement of the new grid, recursively calculate the distance between microelement of the old grid in the balanced binary tree retrieval structure, locate the nearest microelement of the old grid, and establish the index of the nearest microelement of the old grid; According to the index and the old dimensionless flow field data, the estimated dimensionless flow field value of the grid micro-element in the new grid is calculated; The numerical simulation of the new working condition is completed with the estimated dimensionless flow field value, and the aerodynamic data of the aircraft under the new working condition is obtained.

2. The method for multiplexing fluid mechanics flow field data according to claim 1, characterized in that: The original calculated old grid data includes at least the dimensionless position coordinates of each grid element and the minimum bounding box range of the grid block, and also includes one or more of the grid type, grid partitions, topological relationships between partitions, sequence numbers of each element, and connection relationships between each element.

3. The method for multiplexing fluid mechanics flow field data according to claim 1, characterized in that: The old dimensionless flow field data includes one or more of gas pressure, temperature and density.

4. The method for multiplexing fluid mechanics flow field data according to claim 1, characterized in that: The parameters of the new grid include at least the grid type, grid partition, grid block number, dimensionless position coordinates of the grid element and the minimum bounding box range of the grid block, and also include one or more of the topological relationship between the grid partitions, the sequence number of each element, the type of each element and the connection relationship between each element.

5. The method for multiplexing fluid mechanics flow field data according to claim 1, characterized in that: The estimated dimensionless flow field values ​​of the grid elements in the new grid are calculated to include: The spatial dimensionless flow field data corresponding to the index position is read from the dimensionless flow field data, and assigned to the corresponding spatial point position of the new grid element. The assignment method includes direct assignment or distance-based spatial interpolation.

6. The method for multiplexing fluid mechanics flow field data according to claim 1, characterized in that: When calculating the estimated dimensionless flow field value of the grid element in the new grid according to the index and the old dimensionless flow field data, the old dimensionless flow field data is read by parallel reading and writing technology. When reading the data, only the old grid data and dimensionless flow field data that have an intersection relationship with the storage grid in the local process are read.

7. The method for multiplexing fluid mechanics flow field data according to claim 1, characterized in that: Determining whether the new grid and the old grid have an intersection relationship in space specifically includes: The minimum bounding box includes the minimum and maximum data of all the infinitesimal elements of the grid block along the coordinate direction in the Cartesian coordinate axis. By comparing the minimum bounding boxes of the new and old grid blocks, the grid blocks that do not touch or intersect in the geometric space are excluded.

8. The method for multiplexing fluid mechanics flow field data according to claim 1, characterized in that: Methods The HDF5 standard data format was used to store grid data and dimensionless flow field data.

Images