An optimization method and system for numerical simulation of helicopter flow field

Through the combination of SRBT randomization processing, low-rank matrix and improved ACA algorithm, the problem of high computational complexity in numerical simulation of helicopter flow field is solved, efficient numerical simulation of flow field is realized, and the solution speed and accuracy are improved, and the helicopter is used to improve the solution speed and accuracy. Aerodynamic design provides efficient data support.

CN112632874BActive Publication Date: 2024-11-29HANGZHOU DIANZI UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202011632168.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-12-31
Publication Date
2024-11-29
Estimated Expiration
2040-12-31

AI Technical Summary

Technical Problem

The existing numerical simulation method of helicopter flow field is highly complex in large-scale problems, resulting in large time consumption, limiting its application range, and iterating too many times when tile blocks with poor low rank characteristics, affecting efficiency.

Method used

SRBT randomization processing and low-rank matrix application are adopted, combined with tile data layout and improved ACA algorithm, parallel calculations are performed through OpenMP task construct, reducing the number of operations and improving the degree of parallelism, and using iterative refinement algorithm to improve accuracy.

Benefits of technology

It significantly improves the solution speed and accuracy of numerical simulation of flow field, reduces operands, improves calculation speed, and provides efficient data support for the aerodynamic design of helicopters.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112632874B_ABST
    Figure CN112632874B_ABST
Patent Text Reader

Abstract

The present invention discloses an optimization method and system for numerical simulation of helicopter flow field, which can accelerate the numerical simulation analysis of the airflow state during the flight of the helicopter and provide data reference for the aerodynamic design of the helicopter. The system as a whole includes a data acquisition module, a data preprocessing module, a flow field solution module and a grid boundary data update module. The present invention mainly optimizes the performance of the flow field solution module and can improve its solution rate. The main calculation amount of the flow field solution module is in the solution of algebraic equations. Combined with the randomization processing of SRBT and the use of low-rank matrices, the number of operations is reduced to the greatest extent. The present invention greatly improves the solution speed of the flow field while ensuring a certain degree of accuracy, and accelerates the update speed of the numerical simulation of the flow field.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of helicopter aerodynamics, and relates to an optimization method and system based on numerical simulation of helicopter flow field. Background Art

[0002] The aerodynamic force of the whole helicopter is an important means to test the flight ability of the helicopter. There are three main means to study the aerodynamic problem: theoretical analysis (based on the simplified assumption of flow, giving the analytical solution of the problem under study), experimental simulation (using ground experiments as research means) and numerical calculation. The numerical simulation method uses the basic laws (law of conservation of mass, Newton's second law, law of conservation of energy, gas state equation, etc.) and related constitutive equations and state equations to form a partial differential equation group describing the rate of change of air state (density, velocity, pressure) over time, and then the integral and differential terms in the equation are approximately expressed as discrete algebraic forms, so that the control equations in the integral form are converted into algebraic equations, and finally these algebraic equations are solved by computer, thereby obtaining the numerical solution of the flow field in time and space. The present invention mainly improves the performance of the entire numerical simulation by optimizing the solution of algebraic equations. The coefficient matrix of the algebraic equation can be represented by a sparse matrix, and we use the multifrontal method to calculate. In the multi-wavefront method, the calculation of the wavefront matrix is ​​a dense matrix calculation problem, and because the calculation task is divided into the calculation of multiple wavefront matrices, the calculation process involving the wavefront matrix becomes the core step. The direct solver of large linear equations is widely used because of its good numerical robustness, reliability and high availability. However, it also has the disadvantages of high computational complexity and high memory occupancy. This also limits the application scope of the direct solution method in large-scale problems (matrices with hundreds of millions of unknowns). The goal of this technology is to reduce the time consumption of solving algebraic equations while ensuring robustness, computational performance and high availability, which can improve the performance of the entire helicopter flow field numerical simulation. Summary of the invention

[0003] One purpose of the present invention is to solve the above-mentioned problem and to propose an optimization system for numerical simulation of helicopter flow field, which can accelerate the numerical simulation analysis of airflow state during helicopter flight and provide data reference for the aerodynamic design of helicopter. The overall process is divided into the following four modules: data acquisition module, data preprocessing module, flow field solution module and grid boundary data update module. The present invention mainly optimizes the performance of the flow field solution module and can improve its solution rate. The main calculation amount of the flow field solution module is in the solution of algebraic equations. Combining the randomization processing of SRBT (Symmetric Random Butterfly Transformation) and the use of low-rank matrices, the number of operations is reduced to the greatest extent. Under the tiled data layout, the granularity of computing tasks is reduced, the degree of parallelism is improved, low-rank matrices are introduced, and the existing ACA (Adaptive Crossing Approximation) algorithm is improved, which solves the problem of too many calculation iterations when compressing some tiles with poor low-rank characteristics. The number of operations is greatly reduced, and the compression operation can be completed under the minimum low-rank threshold conditions required by the application. Then, the OpenMP task construct is used for parallelization, which greatly improves the computing speed on the shared memory platform. Finally, the iterative refinement algorithm is used to improve the accuracy of the final result, which greatly improves the solution speed of the flow field while ensuring a certain accuracy, and accelerates the update speed of the flow field numerical simulation.

[0004] The system of the present invention comprises a data acquisition module, a data preprocessing module, a flow field solving module and a grid boundary data updating module.

[0005] The data acquisition module is used to obtain the geometric parameters of the helicopter fuselage and the physical parameter information of the airflow during the flight, and the physical parameter information of the airflow includes constructing the incoming flow speed, angle of attack, etc. during the helicopter flight;

[0006] The data preprocessing module is used to construct an airflow grid topology structure, wherein the airflow grid topology structure is grid-divided according to an overlapping grid method, and each grid contains physical parameter information of the airflow at the current position;

[0007] The flow field solving module is used to numerically solve the flow field in each grid of the airflow grid topological structure to obtain flow field data;

[0008] The grid boundary data updating module is used to update the grid nodes of the airflow grid topological structure according to the flow field data obtained by the flow field solving module, so as to achieve the purpose of flow field numerical simulation.

[0009] Another object of the present invention is to provide an optimization method for numerical simulation of helicopter flow field.

[0010] Step (1), obtaining helicopter fuselage geometric parameters and airflow physical parameter information during flight, the airflow physical parameter information including constructing incoming flow velocity, angle of attack, etc. during helicopter flight;

[0011] Step (2), constructing an airflow grid topology structure according to the data obtained in step (1), wherein the airflow grid topology structure is gridded according to an overlapping grid method, and each grid contains physical parameter information of the airflow at the current position;

[0012] Step (3), each grid in the airflow grid topology structure performs independent flow field calculation to obtain grid boundary data of the flow field spatial distribution to provide the values ​​required for flow field update;

[0013] Step (4), based on the grid boundary data of the flow field spatial distribution in step (3), the airflow grid topology structure is updated.

[0014] Step (5), repeat steps (2), (3), and (4) until the flow field converges.

[0015] Another object of the present invention is to provide a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to execute the above method.

[0016] Another object of the present invention is to provide a computing device, comprising a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, the above method is implemented.

[0017] The technical solution provided by the present invention has the following beneficial effects:

[0018] Under the premise of ensuring the accuracy requirements, this method can greatly improve the flow field solution speed, improve the flow field prediction efficiency, and provide data reference for the aerodynamic design of helicopters. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a module architecture diagram of the system of the present invention;

[0020] Figure 2 Flowchart for matrix decomposition of LDLT tiling algorithm;

[0021] Figure 3 (a)-3(b) are column-major data layout and tile-block data layout respectively;

[0022] Figure 4 Decompose the LDLT solution into various tasks in the solver and implement the data correlation in parallel;

[0023] Figure 5(a)-5(b) are the full-rank matrix data format and low-rank matrix data format before and after compression respectively;

[0024] Figure 6 Flowchart for improving the ACA algorithm;

[0025] Figure 7 Flowchart of the mesh boundary data update module. DETAILED DESCRIPTION

[0026] The specific embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings:

[0027] The spatial distribution of the flow field obtained by the optimization method based on the numerical simulation of the helicopter flow field of the present invention provides data support for the optimization design of the aerodynamic shape of the aircraft;

[0028] like Figure 1 The optimization system based on helicopter flow field numerical simulation of the present invention comprises a data acquisition module, a data preprocessing module, a flow field solving module and a grid boundary data updating module.

[0029] The data acquisition module is used to obtain the geometric parameters of the helicopter fuselage and the physical parameter information of the airflow during the flight, and the physical parameter information of the airflow includes constructing the incoming flow speed, angle of attack, etc. during the helicopter flight;

[0030] The data preprocessing module is used to construct an airflow grid topology structure, wherein the airflow grid topology structure is grid-divided according to an overlapping grid method, and each grid contains physical parameter information of the airflow at the current position;

[0031] The flow field solving module is used to numerically solve the flow field in each grid of the airflow grid topological structure to obtain flow field data;

[0032] The grid boundary data updating module is used to update the grid nodes of the airflow grid topological structure according to the flow field data obtained by the flow field solving module, so as to achieve the purpose of flow field numerical simulation.

[0033] The optimization method of the present invention based on helicopter flow field numerical simulation is specifically:

[0034] Step (1), obtaining helicopter fuselage geometric parameters and airflow physical parameter information during flight, the airflow physical parameter information including constructing incoming flow velocity, angle of attack, etc. during helicopter flight;

[0035] Step (2), constructing an airflow grid topology structure according to the data obtained in step (1); wherein the airflow grid topology structure is gridded according to an overlapping grid method, and each grid contains physical parameter information of the airflow at the current position;

[0036] The airflow grid topology is established by an overlapping grid method, specifically:

[0037] S1, dividing the flow field into several regions with overlapping parts;

[0038] S2, generating a grid structure in each area;

[0039] S3, solve each region independently;

[0040] S4. Flow field information is exchanged at the overlapping parts by interpolation between grids.

[0041] Step (3), each grid in the airflow grid topology structure performs independent flow field calculation to obtain grid boundary data of the flow field spatial distribution to provide the values ​​required for flow field update;

[0042] 3-1 Air flow mainly follows four laws: kinematics, dynamics, thermodynamics, and physical and chemical properties of gases, including the law of conservation of mass, Newton's second law, the law of conservation of energy, the entropy equation, the gas state equation, etc. Based on the above basic laws and related constitutive equations and state equations, a set of partial differential equations describing the rate of change of air state (density, velocity, pressure) over time can be formed, namely the Navier-Stocks equations, referred to as NS equations. A three-dimensional compressible Reynolds averaged NS equation is established on each grid in the airflow grid topology:

[0043]

[0044] Where W is a conserved variable, F(W) and F v are convective flux and viscous flux respectively. The partial differential equations on each grid and the final result of the above NS equations can be approximated as a linear algebraic equation system, and this linear equation system is:

[0045] Ax=b (1)

[0046] The coefficient matrix A has a matrix dimension of n, nb is the block size, and x is the unknown quantity in the equation;

[0047] The coefficient matrix A is composed of data blocks of size nb*nb, with a total of nt*nt data blocks;

[0048] Extract the coefficient matrix A of the substitution equation;

[0049]

[0050] 3-2 The coefficient matrix A is preprocessed randomly by the SRBT method, and the matrix A is r Transformed into tiles;

[0051] A r is the random matrix after randomization of coefficient matrix A:

[0052] A r =UAU T (4)

[0053] Where T represents transpose, U represents random matrix, and after randomization, A r The matrix size is still equal to the original matrix A;

[0054] The random matrix U is a recursive butterfly matrix:

[0055] U=U d ×…×U1 (5)

[0056] Where d represents the depth, which is usually less than or equal to 2, U q (1≤q≤d) represents a block diagonal matrix that generates a random matrix U:

[0057]

[0058] Among them B p Represents a butterfly matrix, whose size is n / 2 q-1 (1≤p≤2 q-1 ), let n1=n / 2 q-1 , then the butterfly matrix of size n1×n1 is defined as:

[0059]

[0060] Where R p and S p are random diagonal matrices and non-singular matrices of size n1 / 2×n1 / 2; R p The value of the element in is determined by e ρ / 10 Generate, ρ is Random values ​​in the range, so the butterfly matrix B p The range of elements in is:

[0061] e -1 / 20 ≤r p ≤e 1 / 20 (8)

[0062] Tile-like means dividing the matrix into grid units and storing them sequentially in units of a tile block, which refines the task granularity and improves the efficiency of cache usage;

[0063] Figure 3 (a) The existing matrix column-first data layout is only stored according to the actual structure of the data, and there is no correlation extracted from the associated data during actual use. The cache is not well reused, resulting in a low cache hit rate when accessing data, thereby reducing access performance. Figure 3(b) is a tile-like data layout. Since the matrix operation is performed in units of a single tile block, the tile-like data layout can improve the cache hit rate and make full use of the cache's fast access characteristics. r from Figure 3 The layout format shown in (a) is transformed into Figure 3 (b) The layout format shown.

[0064] 3-3 Decomposing the tile matrix A using the multi-wavefront method r ,Extract the wavefront matrix;

[0065] The multi-wavefront method reorganizes the operations related to sparse decomposition. This reorganization is based on the structure of the elimination tree. Through this reorganization, the task of decomposing a large sparse matrix is ​​transformed into a series of subtasks, and each subtask only includes a local decomposition of a smaller dense wavefront matrix Q.

[0066]

[0067] 3-4 Use the improved LDLT tile algorithm to perform LDLT decomposition on the wavefront matrix Q, and use OpenMP taskconstruct to parallelize it to obtain the lower triangular matrix L and the diagonal matrix D;

[0068] Figure 2 The whole process of decomposition using the improved LDLT tile algorithm is described. Figure 4 Taking nt=3 as an example, assuming that the matrix is ​​divided into 3×3 blocks, the task relevance and execution order in the outermost loop process during the entire decomposition process. In the figure, the factor step is the following step b), the solve step is the following step c), the compress step is the following step d), and the update step is the following step f). The arrows in the figure indicate the direct sequence dependency of each step, and also explain how to parallelize the decomposition process.

[0069] The LDLT decomposition is to decompose the matrix Q into:

[0070] Q=LDL T (10)

[0071]

[0072] Each element in the L and D matrices represents an nb*nb tile block. T refers to the transposition operation of the matrix.

[0073] The improved LDLT tile algorithm is specifically as follows:

[0074] a) Initialize k = 1;

[0075] b) Using LDLT algorithm to calculate the diagonal block Q of wavefront matrix Q kk Decompose and get the lower triangular matrix L kk and the diagonal matrix D kk :

[0076]

[0077] Where 1≤k≤nt, T represents transposition;

[0078] c) According to the lower triangular matrix L obtained above kk and the diagonal matrix D kk Calculate the matrix L of all tiles in the kth column of the wavefront matrix Q ik ;

[0079] W kk =L kk D kk (15)

[0080]

[0081] Where k≤i≤nt.

[0082] d) Using the improved ACA algorithm to compress the tile blocks;

[0083] Figure 5 (a) and Figure 5 (b) describes the format changes of the data before and after compression. The matrix after low-rank compression has the characteristics of low storage space occupancy and low number of computational operations. For step e), which accounts for the largest proportion of computational workload, compression processing before step e) can greatly improve the computational performance of step e).

[0084] Figure 6 The specific implementation process of the improved ACA algorithm is described in . Each loop will generate a set of vectors (a row vector and a column vector, respectively incorporated into matrices X and Y) according to the traditional method of the ACA algorithm to approximate the value of the entire matrix until the approximate accuracy reaches the low-rank threshold, or the number of loops reaches the set loop threshold, then the loop ends and the compression step ends. The approximate accuracy can be calculated by the second normal form of the difference between the low-rank matrix and the original matrix. When the traditional ACA algorithm encounters a matrix with poor low-rank properties, there will be too many loops, too long operation time, and too many vector groups generated, and the subsequent calculation optimization will not be obvious. Therefore, we set a loop threshold to limit the occurrence of this situation. When the loop reaches the loop threshold, we will stop compression and output the result. After that, the operation involving this matrix block will use the original matrix value instead of the compressed matrix.

[0085] Therefore, the compression algorithm steps are as follows: the lower triangular matrix L of all tile blocks in the kth column of step c) is ik According to the traditional method of ACA algorithm, the matrix L ik Decompose into two low-rank matrices X, Y; if the accuracy reaches the low-rank threshold, or the number of vector groups in the low-rank matrix reaches the loop threshold, the loop ends. Otherwise, a new vector group is generated and added to X and Y to generate new X and Y low-rank matrices.

[0086]

[0087] Among them, k+1≤i≤nt.

[0088] e) Update Q ij

[0089] Determine whether the rank of the compressed low-rank matrices X and Y is greater than or equal to the cyclic threshold in step d). If so, use the original full-rank matrix L ik Update Q ij :

[0090]

[0091] Otherwise, substitute formula (17) into formula (18) and use the low-rank matrix to update Q ij :

[0092]

[0093] Among them, k+1≤i≤nt, k+1≤j≤i

[0094] f) Determine whether k is nt in the current iteration. If so, the decomposition of all tile blocks is completed. If not, update k=k+1 and return to step b).

[0095] 3-5 Obtain the solution to the final system of equations through forward substitution and backward substitution;

[0096] The decomposition result of the wavefront matrix Q obtained from the previous step

[0097] Q=LDL T (10)

[0098] The decomposition result of the original coefficient matrix A is obtained by synthesizing the local decomposition results of a series of wavefront matrices Q.

[0099] The decomposition result of the original coefficient matrix A obtained by subsequent calculation of the multi-wavefront method is

[0100]

[0101] Where L merge and D merge The matrices A, LDL and TThe lower triangular and diagonal matrices of the decomposition result.

[0102] According to formula (4), we know

[0103] Ax=U T b U (21)

[0104] Combining formulas (20)-(21), we can know that:

[0105]

[0106] Obtain the intermediate variable y, and then substitute formula (1) into (22) to obtain the solution x:

[0107] x=Uy (23)

[0108] 3-6 Use one iteration optimization to improve the accuracy of the solution.

[0109] 3-6-1 Calculate the residual r from the solution x obtained from formula (23):

[0110] r=b-Ax (24)

[0111] 3-6-2 Calculate the residual r from the solution x obtained from formula (24):

[0112] Ad=r (25)

[0113] Where d is the correction value.

[0114] 3-6-3 Update correction value

[0115] x'=x+d (26)

[0116] Where m represents the number of iterations.

[0117] Since the previously obtained solution itself has a certain degree of accuracy, and considering that the time cost of performing iterative refinement is relatively high, only one iterative refinement operation is performed, that is, the above steps 3-6-1 to 3-6-3 are only performed once, so x' is the solution.

[0118] The grid boundary data of the spatial distribution of the flow field is obtained according to the updated solution x' after iterative optimization.

[0119] Step (4), updating the airflow grid topology structure according to the grid boundary data of the flow field spatial distribution in step (3); Figure 7 Specifically:

[0120] 4-1 Dynamic mesh generation (mesh node update);

[0121] 4-2 Update geometric data (calculate the normal velocity of the control surface according to the law of geometric conservation);

[0122] 4-3 Overlapping grid relationship is established;

[0123] 4-4 External iterative interpolation between subgrids.

[0124] Step (5): In the calculation of non-stationary fluid, it is necessary to repeat the operations of grid establishment, flow field calculation within the grid, and grid update until the flow field converges.

[0125] Therefore, the purpose of numerical simulation of helicopter flow field can be achieved by updating the data after flow field solution and grid boundary data.

Claims

1. An optimization method based on computational fluid dynamics numerical simulation, characterized in that The method comprises the following steps: Step (1), obtaining helicopter fuselage geometric parameters and airflow physical parameter information during flight; Step (2), constructing an airflow grid topology structure based on the data obtained in step (1); wherein each grid contains airflow physical parameter information at the current position; Step (3), each grid in the airflow grid topology structure performs independent flow field calculation to obtain grid boundary data of the flow field spatial distribution; 3-1 The three-dimensional compressible Reynolds averaged NS equation is established on each grid in the airflow grid topology structure: Where W is a conserved variable, F(W) and F v are the convective flux and the viscous flux, respectively; The partial differential equations on each grid and the final result of the above NS equations can be approximated as a linear algebraic equation system, and this linear equation system is: Ax=b (1) The coefficient matrix A has a matrix dimension of n, nb is the block size, and x is the unknown quantity in the equation; The coefficient matrix A is composed of data blocks of size nb*nb, with a total of nt*nt data blocks; Extract the coefficient matrix A of the substitution equation; 3-2 The coefficient matrix A is preprocessed randomly by the SRBT method, and the matrix A is r Transformed into a tile shape, where A r is the random matrix after the coefficient matrix A is randomized; the SRBT method specifically includes the following operations: THE r =WOW T (4) Where T represents transpose, U represents random matrix, and after randomization, A r The matrix size is still equal to the original matrix A; The random matrix U is a recursive butterfly matrix: U=U d ×…×U1 (5) Where d represents the depth, U q Represents a block diagonal matrix that generates a random matrix U, where 1≤q≤d: Among them B p Represents a butterfly matrix, whose size is n / 2 q-1 , let n1=n / 2 q-1 , where 1≤p≤2 q-1 ; Then the butterfly matrix of size n1×n1 is defined as: Where R p and S p Denote random diagonal matrices and non-singular matrices, respectively, with sizes of n1 / 2×n1 / 2; R p The value of the element in is determined by e ρ / 10 Generate, ρ is Random values ​​in the range, so the butterfly matrix B p The range of elements in is: and -1 / 20 ≤r p ≤e 1 / 20 (8); 3-3 Decomposing the tile matrix A using the multi-wavefront method r , extract the wavefront matrix Q; 3-4 Use the improved LDLT tile algorithm to perform LDLT decomposition on the wavefront matrix Q, and use OpenMP taskconstruct to parallelize it to obtain the lower triangular matrix L and the diagonal matrix D; The LDLT decomposition is to decompose the matrix Q into: Q=LDL T (10) Each element in the L and D matrices represents an nb*nb tile block; T refers to the transposition operation of the matrix; The improved LDLT tile algorithm is specifically as follows: a) Initialize k = 1; b) Using LDLT algorithm to calculate the diagonal block Q of wavefront matrix Q kk Decompose and get the lower triangular matrix L kk and the diagonal matrix D kk : Where 1≤k≤nt, T represents transposition; c) According to the lower triangular matrix L obtained above kk and the diagonal matrix D kk Calculate the matrix L of all tiles in the kth column of the wavefront matrix Q ik ; W kk =L kk D kk (15) Where k≤i≤nt; d) Using the improved ACA algorithm to compress the tile blocks; The lower triangular matrix L of all tiles in the kth column of step c) ik According to the traditional method of ACA algorithm, the matrix L ik Decompose into two low-rank matrices X and Y; if the accuracy reaches the low-rank threshold, or the number of vector groups in the low-rank matrix reaches the loop threshold, the loop ends. Otherwise, a new vector group is generated and added to X and Y to generate new X and Y low-rank matrices; Where k+1≤i≤nt; e) Update Q ij Determine whether the rank of the compressed low-rank matrices X and Y is greater than or equal to the cycle threshold in step d). If so, use the original full-rank matrix L ik Update Q ij : Otherwise, substitute formula (17) into formula (18) and use the low-rank matrix to update Q ij : Where k+1≤i≤nt, k+1≤j≤i; f) Determine whether k is nt in the current iteration. If so, all tile block decomposition is completed. If not, update k=k+1 and return to step b); 3-5 The final solution of equation system (1) is obtained by forward substitution and backward substitution; The decomposition result of the wavefront matrix Q obtained from the previous step Q=LDL T (10) The decomposition result of the original coefficient matrix A is obtained by synthesizing the local decomposition results of a series of wavefront matrices Q; The decomposition result of the original coefficient matrix A obtained by subsequent calculation of the multi-wavefront method is Where L merge and D merge The matrices A, LDL and T The lower triangular matrix and diagonal matrix of the decomposition result; According to formula (4), we know Ax=U T bU ( 21 Combining formulas (20)-(21), we can know that: Obtain the intermediate variable y, and then substitute formula (1) into (22) to obtain the solution x: x=Uy (23) 3-6 Use one-time iterative optimization to improve the accuracy of the solution: 3-6-1 Calculate the residual r from the solution x obtained from formula (23): r=b-Ax (24) 3-6-2 Calculate the residual r from the solution x obtained from formula (24): Ad=r (25) Where d is the correction value; 3-6-3 Update correction value x'=x+d (26) According to the updated solution x' after iterative optimization, the grid boundary data of the flow field spatial distribution is obtained; Step (4), updating the airflow grid topology structure according to the grid boundary data of the flow field spatial distribution obtained in step (3); Step (5), repeat steps (2), (3), and (4) until the flow field converges.

2. The optimization method based on computational fluid dynamics numerical simulation according to claim 1, characterized in that Step (1) The airflow physical parameter information includes the incoming flow velocity and angle of attack during the helicopter flight.

3. An optimization system for implementing the helicopter flow field numerical simulation method according to any one of claims 1 to 2, characterized in that It includes data acquisition module, data preprocessing module, flow field solution module and grid boundary data update module; The data acquisition module is used to obtain the geometric parameters of the helicopter fuselage and the physical parameters of the airflow during flight; The data preprocessing module is used to construct an airflow grid topology structure; wherein the airflow grid topology structure is grid-divided according to an overlapping grid method, and each grid contains airflow physical parameter information at the current position; The flow field solving module is used to numerically solve the flow field in each grid of the airflow grid topological structure to obtain flow field data; The grid boundary data updating module is used to update the grid nodes of the airflow grid topological structure according to the flow field data obtained by the flow field solving module, so as to achieve the purpose of flow field numerical simulation.

4. A computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to execute the method according to any one of claims 1 to 2.

5. A computing device, comprising a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, the method according to any one of claims 1 to 2 is implemented.

Citation Information

Patent Citations

  • Simulation method and system for flight simulation of helicopter

    CN108920811A

  • A supercomputer-based overlap optimization method for parallel CFD computation and communication is proposed

    CN109088776A