New numerical solution method for eigenvalue equation boundary value problem

Through adaptive mesh division and numerical solution methods combined with finite difference method and finite element method, the problem of limited calculation efficiency and accuracy of the eigenvalue equation boundary value problem in the prior art is solved, and efficient solution of complex eigenvalue boundary values ​​is achieved Problem ability.

CN119004887APending Publication Date: 2024-11-22玉林 +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410998587.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

When existing numerical methods deal with complex boundary value problems of eigenvalue equations, especially high-dimensional and nonlinear problems, the calculation efficiency and accuracy are limited by the meshing strategy and are difficult to effectively solve.

Method used

Adaptive mesh division method is used to automatically divide the solution areas according to the gradient or error index of the preliminary solution, and discrete the finite difference method and finite element method in each area, and solve the linear equation system through the eigenvalue solution algorithm, which is suitable for Two-dimensional and three-dimensional problems.

Benefits of technology

It improves computational efficiency and accuracy, can effectively deal with complex eigenvalue boundary problems, especially to provide high-precision solutions in key areas, and reduce unnecessary computational volume, and is suitable for a variety of physical and engineering fields.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004887A_ABST
    Figure CN119004887A_ABST
Patent Text Reader

Abstract

The invention relates to the field of computational mathematics and numerical analysis, and discloses a new numerical solution method for a boundary value problem of a characteristic value equation, which comprises the following steps of: S1, automatically dividing a solution region according to a gradient or error index of a preliminary solution; s2, discretizing in each divided region by applying a finite difference method or a finite element method; and S3, solving the discretized linear equation set to obtain a characteristic value and a corresponding characteristic function. According to the numerical solution method, different numerical technologies and self-adaptive grid strategies are combined, the calculation efficiency can be improved while the solution precision is guaranteed, the method has remarkable advantages for processing a complex eigenvalue boundary value problem, grid refinement can be carried out for a specific region of the problem through self-adaptive grid division, and the method has the advantages of being simple in structure, convenient to operate and the like. Therefore, higher numerical solution precision is obtained in a key area.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of computational mathematics and numerical analysis, and in particular to a new numerical solution method for boundary value problems of eigenvalue equations. Background Art

[0002] The solution of boundary value problems of eigenvalue equations is crucial in many physical and engineering problems. These problems involve a wide range of fields, including quantum mechanics, vibration analysis, stability analysis, heat conduction, fluid dynamics, etc. Eigenvalue problems usually involve solving a special type of differential equation, namely the eigenvalue differential equation, whose solution consists of a set of eigenvalues ​​and corresponding eigenfunctions. In practical applications, these problems usually involve complex geometries and nonlinear physical properties, so numerical methods are needed to solve them.

[0003] Existing numerical methods mainly include finite difference method, finite element method and finite volume method. These methods all require dividing the solution area into a series of small areas or grids, and then performing approximate solutions in each small area. However, the computational efficiency and accuracy of these methods are often limited by the quality of the grid and the division strategy. For example, a grid that is too coarse may lead to insufficient solution accuracy, while a grid that is too fine may result in excessive calculations and low computational efficiency. In addition, existing numerical methods are often difficult to handle for complex nonlinear eigenvalue problems and high-dimensional problems. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a new numerical solution method for the boundary value problem of eigenvalue equations. The method first automatically divides the solution region according to the gradient or error index of the preliminary solution, then discretizes each divided region using the finite difference method or the finite element method, and finally solves the discretized linear equations to obtain the eigenvalues ​​and corresponding characteristic functions. In addition, the method can also be extended to solve nonlinear eigenvalue problems and three-dimensional problems.

[0005] To achieve the above object, the present invention is implemented by the following technical scheme: a new numerical solution method for the boundary value problem of the eigenvalue equation, comprising the following steps:

[0006] Automatically divide the solution area according to the gradient or error index of the preliminary solution;

[0007] Apply finite difference method or finite element method to discretize in each divided area;

[0008] The discretized linear equations are solved to obtain the eigenvalues ​​and the corresponding eigenfunctions.

[0009] Preferably, the step of automatically dividing the solution area includes:

[0010] A preliminary solution was obtained using a coarse grid and simple numerical methods;

[0011] Identify areas that require a finer mesh based on the gradient or error metrics of the preliminary solution;

[0012] Iterate until the predetermined accuracy requirement is met.

[0013] Preferably, the discretization step comprises:

[0014] For the finite difference method region, a five-point difference scheme is applied to approximate the Laplace operator;

[0015] For the finite element method region, bilinear basis functions are selected for weighted residual method discretization.

[0016] Preferably, the step of solving the linear equations comprises:

[0017] Combine the discrete equations in the finite difference method region and the finite element method region to construct the total linear equation system;

[0018] Apply the eigenvalue solving algorithm to solve the total linear equations.

[0019] Preferably, the eigenvalue solution algorithm is a power method or a Jacobi method.

[0020] Preferably, the eigenvalue equation boundary value problem is a two-dimensional problem.

[0021] Preferably, the method is also used to solve the boundary value problem of nonlinear eigenvalue equation, including:

[0022] Linearize nonlinear problems using Newton or quasi-Newton methods;

[0023] The linearized problem is solved using the method described in each iteration.

[0024] Preferably, the method is also used to solve the boundary value problem of a three-dimensional eigenvalue equation, comprising:

[0025] Extend the finite difference method and finite element method from two dimensions to three dimensions;

[0026] Handle larger systems of linear equations, including using parallel computing or efficient iterative methods.

[0027] The present invention also provides a numerical solution device for a boundary value problem of an eigenvalue equation, comprising:

[0028] Adaptive region division module, used to automatically divide the solution region;

[0029] A discretization module for applying finite difference method or finite element method for discretization in each region;

[0030] The solving module is used to solve the discretized linear equations to obtain the eigenvalues ​​and corresponding eigenfunctions.

[0031] The present invention also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the above method is implemented.

[0032] The present invention also provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the method described above is implemented.

[0033] The present invention provides a new numerical solution method for the boundary value problem of the eigenvalue equation. It has the following beneficial effects:

[0034] 1. The numerical solution method of the present invention can improve the computational efficiency while ensuring the solution accuracy by combining different numerical techniques and adaptive grid strategies, and has significant advantages in processing complex eigenvalue and boundary value problems.

[0035] 2. The present invention can refine the grid for specific areas of the problem (such as areas with large solution gradients or high error indicators) through adaptive grid division, thereby obtaining higher numerical solution accuracy in key areas.

[0036] 3. The present invention maintains a coarser grid in areas where the solution does not change significantly, thereby reducing unnecessary calculations and improving overall calculation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0038] Figure 2 It is a schematic diagram of the structure of the device of the present invention;

[0039] Figure 3 It is a schematic diagram of the computer device structure of the present invention.

[0040] Among them, 10, adaptive region division module; 20, discretization module; 30, solution module; 40, computer equipment; 41, processor; 42, memory; 43, storage medium. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] Please see attached Figure 1 , an embodiment of the present invention provides a new numerical solution method for the boundary value problem of eigenvalue equations, which mainly solves the eigenvalue problems encountered in the fields of multi-physics fields, structural mechanics, quantum mechanics, etc. Traditional numerical solution methods often have fixed grid divisions, which are not flexible enough, and cannot adapt to the characteristics of different regions, resulting in waste of computing resources and low computing efficiency. The present invention improves the accuracy and efficiency of the solution through adaptive region division and the strategy of combining different numerical methods.

[0043] The method specifically comprises the following steps:

[0044] S1. Adaptive region division

[0045] In this step, a rough grid and simple numerical methods are first used to obtain a preliminary solution to the problem. Based on the gradient or error estimate of the preliminary solution, the area that needs a more refined solution is automatically identified. For example, for areas with large gradients or error indicators exceeding a preset threshold, they are divided into smaller sub-areas, and a finer grid is used in these sub-areas. This adaptive partitioning can dynamically adjust the grid density according to the characteristics of the problem, avoiding over-refining the grid in areas where the characteristics do not change much, thereby saving computing resources and improving computational efficiency.

[0046] In this embodiment, the step of automatically dividing the solution region is implemented through a series of iterative processes to ensure that a high-precision numerical solution is obtained in the key region of the eigenvalue problem.

[0047] 1. Obtaining the preliminary solution

[0048] First, set a coarse grid division and cover the solution area with an initial grid system. In this system, the function value at each grid point is unknown. Use simple numerical methods, such as finite difference method, to solve the initial approximate solution of the differential equation. For example, for a two-dimensional Laplace equation, its discretized form can be expressed as:

[0049]

[0050] Among them, u i,j is the function value at the grid point (i, j), and Δx and Δy are the grid spacing in the x and y directions.

[0051] 2. Calculation of gradient or error index

[0052] Once a preliminary solution is obtained, the gradient or error metric of the solution at each grid point can be calculated. The gradient can be estimated using the central difference formula:

[0053]

[0054] The error index can be estimated by comparing the difference between the current solution and the exact solution. If the exact solution is not known, the rate of change or residual of the solution can be used as an indicator:

[0055]

[0056] Among them, E i,j represents the error estimate at the grid point (i,j).

[0057] 3. Adaptive meshing

[0058] Based on the calculated gradient or error index, determine the area that needs mesh refinement. A common strategy is to set a threshold ∈, and refine the mesh around those grid points with gradients greater than ∈ or error indicators exceeding ∈. Mesh refinement can be achieved by reducing the values ​​of Δx and Δy, thereby providing higher local resolution in the next iteration.

[0059]

[0060] For the selected area, new grid points will be added and the equations will be re-solved on the new grid system to obtain an updated solution.

[0061] 4. Iteration process

[0062] This adaptive meshing and solving process is iterative. After each iteration, the gradient or error index is re-evaluated and a decision is made as to whether further mesh refinement is needed. This process continues until the error index for the entire solution region is below the preset accuracy requirement ∈, or the predetermined number of iterations is reached.

[0063]

[0064] or

[0065] Iterations > Predetermined number

[0066] Through the above steps, the present invention can not only provide high-precision numerical solutions in key areas, but also effectively control the computational cost, making the solution process more efficient. In addition, the adaptability of the method enables it to cope with various complex boundary conditions and nonlinear problems, and has a wide range of application potentials.

[0067] S2. Discretization solution area

[0068] In each divided area, the appropriate numerical method is selected for discretization according to the characteristics of the area. In areas where the characteristics change gently, the finite difference method can be used; while in areas where the characteristics change drastically, the finite element method is used. The main advantages of the finite difference method are simplicity and efficiency, and it is suitable for regular grids; the finite element method is more flexible in dealing with complex boundaries and non-uniform material characteristics. By combining these two methods, the solution speed can be improved while ensuring accuracy.

[0069] The discretization step in this embodiment involves two different numerical methods: finite difference method (FDM) and finite element method (FEM). These two methods are respectively applicable to different regions and perform discretization processing according to their respective characteristics.

[0070] 1. Five-point difference scheme in the finite difference method area

[0071] The finite difference method is a technique for numerically solving partial differential equations. In particular, the five-point difference scheme is a common method for approximating the two-dimensional Laplace operator. On a uniform grid, the Laplace operator The discrete approximation at point (i, j) can be expressed as:

[0072]

[0073] Here Δx is the spacing of the grid in the x and y directions (assuming a square grid). This equation is due to the fact that at a point (i, j), the discrete representation of the Laplacian operator involves the function value of the four neighbors to the north, south, east, and west of that point.

[0074] 2. Weighted residual discretization of finite element method region

[0075] Finite element method is another numerical solution technique that approximately solves the equations by dividing the solution region into a finite number of sub-regions (elements) and selecting appropriate basis functions. For two-dimensional problems, the commonly used basis functions are bilinear basis functions, which are linear on each element. In the finite element method, the equations are solved by minimizing the weighted average of the residuals in the entire solution region.

[0076] Let u(x,y) be the Laplace equation In the finite element method, we find an approximate solution u h (x,y), which is the basis function φ i Linear combination of (x,y):

[0077]

[0078] where c i is the coefficient to be determined, and N is the number of basis functions. The weighted residual method requires the residual The weighted integral over each basis function is zero:

[0079]

[0080] will u h Substituting the expressions of and R(x,y) into the above equations and applying Green's formula, we can get a linear equation system to solve the coefficient c i :

[0081]

[0082] This system of equations forms the basis of the finite element method, and its solution provides u h The coefficient c of (x,y) i , which can then be used to approximate the solution to the original equation.

[0083] Through the discretization steps of the above two methods, the approximation of the numerical solution can be obtained in the appropriate region. The finite difference method is widely used because of its simplicity and efficiency in processing on regular grids, while the finite element method is more flexible and accurate in processing complex geometries and boundary conditions. By properly combining these two methods, the computational efficiency can be improved while ensuring the accuracy of the solution.

[0084] S3, solving the discretized linear equations to obtain eigenvalues ​​and corresponding eigenfunctions;

[0085] Solve the discretized equations to obtain the eigenvalues ​​and corresponding eigenfunctions of the problem. This step can use a direct solver or an iterative solver, such as the power method, inverse iteration method, Lanczos method, etc. The most suitable algorithm is selected according to the scale and characteristics of the problem. During the solution process, the interface conditions of different regions need to be processed to ensure that the solution of the entire solution area is continuous and consistent.

[0086] In this embodiment, the step of solving the linear equations is the core of the entire numerical solution method. This step involves fusing the discrete equations in the finite difference method (FDM) region and the finite element method (FEM) region into a total linear equation system, and then applying the eigenvalue solution algorithm to solve the equation system.

[0087] 1. Construct the total linear equations

[0088] First, we need to combine the linear equations obtained by FDM and FEM discretization. Suppose the linear equations obtained by finite difference method are The linear equations obtained by the finite element method are: Among them A FDM and A FEM are the coefficient matrices constructed by the FDM and FEM methods, respectively. is the unknown eigenvector, and is the source term vector.

[0089] At the interface between two regions, the continuity of the solution needs to be ensured, which is usually achieved by introducing additional constraints. Assume that the conditions at the interface are where C is the constraint matrix, is the interface condition vector. Then, the total linear equations can be expressed as:

[0090]

[0091] here, and are the solution vectors for the FDM and FEM regions, respectively.

[0092] 2. Eigenvalue solution algorithm

[0093] In order to solve the eigenvalues ​​and eigenvectors of the total linear equations, the power method or the Jacobi method can be used.

[0094] Power Method

[0095] The power method is an iterative method for computing the principal eigenvalue and corresponding eigenvector of a matrix. For a given matrix A, the eigenvector of the initial guess is The power method approximates the principal eigenvalue λ and the corresponding eigenvector by the following iterative process

[0096]

[0097] Where k is the number of iterations and ∥·∥ represents the norm of the vector.

[0098] Jacobi Method

[0099] The Jacobi method is a method for finding eigenvalues ​​and eigenvectors of a symmetric matrix. For a given symmetric matrix A, the Jacobi method gradually transforms A into a diagonal matrix through a series of rotation operations, and the diagonal elements are the eigenvalues. The rotation matrix J in each step is selected to eliminate a pair of off-diagonal elements A of A. qp and A qp :

[0100] A (k+1) =J T A (k) J

[0101] Among them A (0) =A, J is the rotation matrix, and k is the number of iterations. This process is repeated until A (k)The off-diagonal elements of are small enough to be considered as (k) is a diagonal matrix.

[0102] Through the above two methods, the eigenvalues ​​and eigenvectors of the total linear equations can be obtained, thereby solving the approximate numerical solution of the eigenvalue problem. The power method is suitable for calculating the maximum eigenvalue, while the Jacobi rule is suitable for calculating all eigenvalues ​​and eigenvectors of a symmetric matrix. In practical applications, the most appropriate algorithm can be selected according to the specific characteristics of the problem.

[0103] In summary, the new numerical solution method for the boundary value problem of the eigenvalue equation of the present invention can refine the grid for specific areas of the problem (such as areas with large solution gradients or high error indicators) through adaptive grid division, thereby obtaining higher numerical solution accuracy in key areas; and, in areas where the solution changes are not significant, a coarser grid is maintained, thereby reducing unnecessary calculations and improving the overall calculation efficiency; at the same time, the discretization strategies of the finite difference method and the finite element method are combined, so that the method can flexibly respond to different problem characteristics, such as complex geometric shapes and boundary conditions; the method is not limited to a specific type of partial differential equations, and can be applied to various types of eigenvalue boundary value problems, such as fluid mechanics, solid mechanics, electromagnetism and other fields.

[0104] In general, the numerical solution method of the present invention can improve the computational efficiency while ensuring the solution accuracy by combining different numerical techniques and adaptive grid strategies, and has significant advantages in processing complex eigenvalue and boundary value problems.

[0105] In the aforementioned implementation, it has been discussed in detail how to apply the finite difference method (FDM) and the finite element method (FEM) and the eigenvalue solution algorithm to solve the two-dimensional linear eigenvalue boundary value problem. However, in practical applications, nonlinear eigenvalue problems are often encountered. These problems are more complicated due to their nonlinear characteristics and cannot be solved directly using linear methods. Next, this embodiment will expand the aforementioned implementation to introduce how to apply these methods to the solution of nonlinear eigenvalue boundary value problems.

[0106] When solving nonlinear eigenvalue boundary value problems, the first challenge is that nonlinear equations usually do not have closed-form analytical solutions, so numerical methods are needed to solve them. A typical example of a nonlinear problem is the nonlinear Schrödinger equation, which has a wide range of applications in fields such as quantum physics and optics. This equation contains one or more nonlinear terms, which make the solution of the problem more complicated.

[0107] In order to solve this type of problem, this embodiment linearizes the nonlinear problem in an iterative manner, and then solves the linearized problem in each iteration. Specifically, the following steps can be used:

[0108] 1. Linearization of nonlinear problems

[0109] The initial step is to transform the nonlinear eigenvalue problem into a series of linear problems. This is usually done by applying Newton's method or quasi-Newton's method. For example, consider the nonlinear eigenvalue problem:

[0110]

[0111] Where F represents an operator containing nonlinear terms. Using Newton's method to iterate, a linearized problem needs to be solved in each iteration, which can be expressed as:

[0112]

[0113] Among them J F is F with respect to λ and The Jacobian matrix of is the solution for the kth iteration.

[0114] 2. Iterative solution of linearized problem

[0115] In each iteration, the linear eigenvalue solving algorithm (such as the power method or the Jacobi method) in the above-mentioned implementation is used to solve the linearized eigenvalue problem. The form of these linear problems can be written as:

[0116]

[0117] Among them A (k) and B (k) Based on the current iterative solution Constructed matrix.

[0118] 3. Update the solution and check convergence

[0119] After each iteration, the estimated values ​​of eigenvalues ​​and eigenvectors are updated and checked to see if the convergence condition is met. If the update amount of the solution is less than a predetermined threshold or the maximum number of iterations is reached, the iteration process stops.

[0120] Through the above steps, the solution of the nonlinear eigenvalue problem can be gradually approximated. This method not only retains the efficiency and accuracy of the previous implementation, but also enables the processing of nonlinear eigenvalue problems by introducing iterative linearization technology, which is crucial for understanding and predicting the behavior of complex systems.

[0121] In the foregoing embodiments, it has been discussed how to use the finite difference method (FDM) and the finite element method (FEM) and the eigenvalue solution algorithm to solve the two-dimensional eigenvalue boundary value problem. The application of these methods to two-dimensional problems provides a solid foundation, which enables these techniques to be further extended to more complex three-dimensional eigenvalue equation boundary value problems.

[0122] Three-dimensional eigenvalue and boundary value problems are of great importance in many areas of science and engineering, for example in simulations of computational fluid dynamics, solid mechanics, and electromagnetic fields. In these problems, the equations that need to be solved are usually of higher complexity, thus requiring more advanced numerical methods and computational techniques.

[0123] 1. Expanding from 2D to 3D

[0124] Extending the finite difference and finite element methods from two dimensions to three dimensions means constructing meshes in three dimensions and applying appropriate discretization techniques on those meshes. For the finite difference method, this involves applying a difference scheme in three spatial dimensions. For example, a five-point difference scheme for the three-dimensional Laplace equation would be extended to a seven-point difference scheme:

[0125]

[0126] For the finite element method, it is necessary to define three-dimensional elements (such as tetrahedrons or hexahedrons) and the corresponding basis functions in order to perform integration and assemble the global stiffness matrix over the two-dimensional domain.

[0127] 2. Handling larger linear equations

[0128] Three-dimensional problems often involve larger systems of linear equations, which require the use of more efficient solvers and computational techniques. These include:

[0129] Iterative methods: such as the conjugate gradient method (CG) or the generalized minimum residual method (GMRES), which are suitable for solving large-scale sparse linear systems, especially when direct solution methods become impractical.

[0130] Parallel computing: Utilizing multi-processor or multi-core computing resources can significantly speed up computations. This involves breaking the problem into sub-problems, each of which is solved in parallel on a different computing core.

[0131] Preconditioner: To increase the convergence rate of iterative methods, preconditioning techniques can be used to improve the condition number of the coefficient matrix.

[0132] Through the above steps, the three-dimensional eigenvalue boundary value problem can be effectively solved. The application of these techniques not only enables the processing of higher-dimensional problems, but also copes with the larger-scale computational challenges brought about by them.

[0133] The numerical solution device for the eigenvalue equation boundary value problem described below and the new numerical solution method for the eigenvalue equation boundary value problem described above can be referenced to each other.

[0134] Please see attached Figure 2The present invention also provides a numerical solution device for the boundary value problem of the eigenvalue equation, comprising:

[0135] An adaptive region division module 10, used for automatically dividing the solution region;

[0136] A discretization module 20, used for applying a finite difference method or a finite element method to discretize in each region;

[0137] The solution module 30 is used to solve the discretized linear equations to obtain eigenvalues ​​and corresponding eigenfunctions.

[0138] The device of this embodiment can be used to execute the above method embodiment, and its principles and technical effects are similar, which will not be repeated here.

[0139] Please see attached Figure 3 The present invention further provides a computer device 40, comprising: a processor 41 and a memory 42, wherein the memory 42 stores a computer program executable by the processor, and when the computer program is executed by the processor, the above method is executed.

[0140] The present invention further provides a storage medium 43 on which a computer program is stored. When the computer program is run by the processor 41, the above method is executed.

[0141] Among them, the storage medium 43 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable red-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, disk or optical disk.

[0142] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A new numerical solution method for boundary value problems of eigenvalue equations, characterized in that: The following steps are involved: Automatically divide the solution area according to the gradient or error index of the preliminary solution; Apply finite difference method or finite element method to discretize in each divided area; The discretized linear equations are solved to obtain the eigenvalues ​​and the corresponding eigenfunctions.

2. A new numerical solution method for the boundary value problem of eigenvalue equation according to claim 1, characterized in that: The step of automatically dividing the solution area includes: A preliminary solution was obtained using a coarse grid and simple numerical methods; Identify areas that require a finer mesh based on the gradient or error metrics of the preliminary solution; Iterate until the predetermined accuracy requirement is met.

3. A new numerical solution method for the boundary value problem of eigenvalue equation according to claim 1, characterized in that: The discretization step comprises: For the finite difference method region, a five-point difference scheme is applied to approximate the Laplace operator; For the finite element method region, bilinear basis functions are selected for weighted residual method discretization.

4. A new numerical solution method for a boundary value problem of an eigenvalue equation according to claim 1, characterized in that: The steps of solving the linear equations include: Combine the discrete equations in the finite difference method region and the finite element method region to construct the total linear equation system; Apply the eigenvalue solving algorithm to solve the total linear equations.

5. A new numerical solution method for the boundary value problem of eigenvalue equation according to claim 4, characterized in that: The eigenvalue solution algorithm is the power method or the Jacobi method.

6. A new numerical solution method for a boundary value problem of an eigenvalue equation according to claim 1, characterized in that: The eigenvalue equation boundary value problem is a two-dimensional problem.

7. A new numerical solution method for the boundary value problem of eigenvalue equation according to claim 6, characterized in that: The method is also used to solve the boundary value problem of nonlinear eigenvalue equation, including: Linearize nonlinear problems using Newton or quasi-Newton methods; The linearized problem is solved using the method described in each iteration.

8. A new numerical solution method for the boundary value problem of eigenvalue equation according to claim 6, characterized in that: The method is also used to solve the boundary value problem of three-dimensional eigenvalue equation, including: Extend the finite difference method and finite element method from two dimensions to three dimensions; Handle larger systems of linear equations, including using parallel computing or efficient iterative methods.

9. A numerical solution device for boundary value problems of eigenvalue equations, based on the solution method according to any one of claims 1 to 8, characterized in that: include: Adaptive region division module, used to automatically divide the solution region; A discretization module for applying finite difference method or finite element method for discretization in each region; The solving module is used to solve the discretized linear equations to obtain the eigenvalues ​​and corresponding eigenfunctions.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.