Spectrum library optimization method and system based on sensitivity analysis

Abstract

The invention discloses a spectrum library optimization method and system based on sensitivity analysis, and belongs to the field of semiconductor optical measurement, and the optimization method comprises the steps: obtaining a simulation spectrum; carrying out sensitivity analysis on preset library building parameters based on the simulated spectrum, and obtaining the sensitivity of each library building parameter; obtaining a parameter value range of each library building parameter, and obtaining a division weight of each library building parameter based on the sensitivity of each library building parameter and the parameter value range of each library building parameter; calculating the grid division number of each library building parameter based on the division weight of each library building parameter; grid points are obtained according to the grid division number of each library building parameter, each grid point corresponds to a set of geometric parameters, and a parameter network is obtained; and geometric parameters corresponding to each grid point in the parameter network are simulated through a Fourier modal method, and an optimized spectrum library is generated. The spectrum library sample obtained through the optimization method is balanced in distribution and small in deviation, and the precision and stability of the neural network model are improved.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor optical measurement, and in particular to a spectral library optimization method and system based on sensitivity analysis. Background Art

[0002] In high-density integrated circuit manufacturing, nanoscale metrology is key to ensuring device performance. Spectroscopic ellipsometry, a key nanoscale metrology technique, is widely used to characterize critical dimensions and optical constants of semiconductor chip structures due to its surface sensitivity, non-destructiveness, and non-invasiveness. It is based on the change in polarization state of linearly polarized light after reflection from a sample. By building a model and calculating a simulated spectrum using the Fourier modal method (FMM) and similar algorithms, the desired result is then fitted to the experimental spectrum.

[0003] Traditional optical geometry measurement methods rely on complex electromagnetic calculations. While accurate, they are computationally expensive and slow. To accelerate the measurement process, neural network models are being used to replace traditional calculations, learning the mapping relationship between parameters and spectra using a pre-generated spectral library. However, the sampling distribution of the spectral library relies on the operator's experience, resulting in uneven sample distribution and large deviations, which in turn affects the accuracy and stability of the neural network model. Summary of the Invention

[0004] This proposal aims to provide a spectral library optimization method and system based on sensitivity analysis to solve the above technical problems, so that the optimized spectral library samples are evenly distributed and have small deviations, thereby improving the accuracy and stability of the neural network model.

[0005] To solve the above problems, the present invention provides a spectral library optimization method and system based on sensitivity analysis, comprising the following steps:

[0006] Get simulated spectra;

[0007] Perform sensitivity analysis on the preset library construction parameters based on simulated spectra to obtain the sensitivity of each library construction parameter;

[0008] Obtain the parameter value range of each library construction parameter, and obtain the partition weight of each library construction parameter based on the sensitivity of each library construction parameter and the parameter value range of each library construction parameter;

[0009] Calculate the number of grid divisions for each database construction parameter based on the division weight of each database construction parameter;

[0010] Obtain grid points based on the number of grid divisions for each database parameter. Each grid point corresponds to a set of geometric parameters to obtain a parameter network.

[0011] The geometric parameters corresponding to each grid point in the parameter network are simulated by the Fourier modal method to generate an optimized spectral library.

[0012] In the above scheme, by obtaining simulated spectra, sensitivity analysis is performed on the preset library construction parameters. That is, the sensitivity of each library construction parameter is calculated by calculating the changes in the library construction parameters to the spectral output. The partition weight of each library construction parameter is obtained based on the parameter value range and sensitivity of each library construction parameter, and the number of grid divisions for each library construction parameter is calculated. Grid points are obtained based on the number of grid divisions for each library construction parameter, where a set of geometric parameters corresponding to each grid point constitutes a parameter network. Finally, the geometric parameters corresponding to each grid point in the parameter network are simulated using the Fourier modal method to generate an optimized spectral library. The obtained optimized spectral library can then be used as a training neural network. The spectrum sampled by the above scheme is evenly distributed and has a small sample deviation, which greatly improves the accuracy and stability of the neural network model.

[0013] Furthermore, the sensitivity analysis of the preset library construction parameters based on the simulated spectrum to obtain the sensitivity of each library construction parameter includes:

[0014] The sensitivity of each library construction parameter is calculated using the partial derivative finite difference method, which is expressed as:

[0015]

[0016] Where F_λ represents the simulated spectrum at the λth wavelength, p_i represents the process expected value of the i-th library construction parameter, Δp_i represents the parameter perturbation step, N represents the wavelength index, and paraSense[i] represents the sensitivity of the i-th library construction parameter.

[0017] In the above scheme, the finite difference method is used to calculate partial derivatives to determine the impact of changes in library parameters on optical measurement results. Specifically, the sensitivity of each library parameter, paraSense[i], is obtained by simulating the output change of the spectrum, providing an important basis for subsequent grid division. In addition to calculating the maximum value of the output change of the simulated spectrum as in the above scheme, the sensitivity of each library parameter can also be obtained by calculating the average value or other variations. Compared to the existing method of manually setting the sampling interval, the above scheme integrates sensitivity analysis into automatic grid division, achieving dynamic adjustment of sampling density and improving accuracy.

[0018] Further, obtaining the parameter value range of each library construction parameter, and obtaining the partitioning weight of each library construction parameter based on the sensitivity of each library construction parameter and the parameter value range of each library construction parameter, specifically includes:

[0019] According to the database construction parameters, calculate the value range of each database construction parameter paraRange[i], which represents the value range of the i-th database construction parameter;

[0020] Multiply the sensitivity of each library construction parameter by the parameter value range of each library construction parameter to obtain the partition weight of each library construction parameter, which is specifically expressed as: paraWeight[i]=paraSense[i]×paraRange[i];

[0021] Where paraWeight[i] represents the partition weight of the i-th database construction parameter, paraSense[i] represents the sensitivity of the i-th database construction parameter, and paraRange[i] represents the parameter value range of the i-th database construction parameter.

[0022] The above scheme calculates the partitioning weight of each library construction parameter and assigns higher weights to more representative library construction parameter data, so that this data plays a greater role in the subsequent steps, thereby improving the accuracy of the neural network training set.

[0023] Furthermore, the calculation of the number of grid divisions for each database construction parameter based on the division weight of each database construction parameter specifically includes:

[0024] Normalize each library construction parameter;

[0025] The maximum and minimum weight values ​​of the database construction parameter division weights are normalized and expressed as:

[0026]

[0027] Among them, normParaWeight[i] is the partition weight of the i-th database construction parameter after normalization, paraWeightMax represents the maximum value of the partition weights of all database construction parameters, and paraWeightMin represents the minimum value of the partition weights of all database construction parameters.

[0028] This approach combines sensitivity analysis with weighted normalization, assigning higher weights to highly sensitive library parameters, making them more important in subsequent calculations. This dynamically adjusts the sampling density for subsequent automatic meshing, providing greater accuracy and stability. Furthermore, this approach is particularly well-suited for ellipsometry, a technique commonly used to characterize critical dimensions and optical constants of semiconductor chips.

[0029] Furthermore, the calculation of the number of grid divisions for each database construction parameter based on the division weight of each database construction parameter specifically includes:

[0030] Calculate the base value base used for preliminary allocation of grid numbers, expressed as:

[0031]

[0032] Where LibCountUser represents the number of preset spectral libraries, and m represents the number of library construction parameters;

[0033] The number of grid divisions for each database construction parameter is calculated based on the normalized division weight and basic value, which can be expressed as follows:

[0034] Grid[i]=normParaWeight[i]×base;

[0035] Among them, Grid[i] represents the number of grid divisions of the i-th database construction parameter.

[0036] This approach uses the user-specified number of spectral libraries and the number of library construction parameters to determine a base value for initial grid number allocation. The base value then calculates the optimal number of grid divisions for each library construction parameter based on the normalized division weights and the base value. By comprehensively considering the sensitivity, value range, and division weight of each library construction parameter, the automatically calculated optimal number of grid divisions avoids the problem of irrational grid divisions caused by manual settings, effectively improving computational efficiency and accuracy.

[0037] Furthermore, the grid points are obtained according to the number of grid divisions of each database construction parameter, each grid point corresponds to a set of geometric parameters, and the parameter network is obtained, which specifically includes:

[0038] The corresponding grid points are calculated based on the number of grid divisions for each parameter. Each grid point corresponds to a set of geometric parameters, expressed as:

[0039]

[0040] Where paraMin[i] represents the minimum value of the i-th database construction parameter, Grid[i] represents the number of grid divisions for the i-th database construction parameter, j represents any positive integer less than Grid[i], and m represents the number of database construction parameters.

[0041] The geometric parameters corresponding to each grid point in the above scheme can form a parameter network. Then, based on the parameter network sampling, the spectral library of the geometric parameters corresponding to the grid point is simulated by the Fourier modal method to generate an optimized neural network training set, thereby realizing an automatic optimization scheme based on sensitivity analysis, ensuring that the sampling points are concentrated in the effective area and improving accuracy.

[0042] The present invention also provides a spectral library optimization system based on sensitivity analysis, comprising: a measurement result acquisition module for acquiring simulated spectra; a sensitivity analysis module for performing sensitivity analysis on preset library construction parameters based on the simulated spectra to obtain the sensitivity of each library construction parameter; a weight division module for obtaining the parameter value range of each library construction parameter, and obtaining the division weight of each library construction parameter based on the sensitivity of each library construction parameter and the parameter value range of each library construction parameter; a grid division module for calculating the number of grid divisions for each library construction parameter based on the division weight of each library construction parameter; a parameter network acquisition module for obtaining grid points according to the number of grid divisions for each library construction parameter, each grid point corresponding to a set of geometric parameters, to obtain a parameter network; and a spectral library optimization module for simulating the geometric parameters corresponding to each grid point in the parameter network by the Fourier modal method to generate an optimized spectral library.

[0043] The system architecture proposed in the above scheme is simple and easy to implement, and can be used to implement a spectral library optimization method and system based on sensitivity analysis as described above. The simulated spectrum is obtained through the measurement result acquisition module; the sensitivity analysis module performs sensitivity analysis on the preset library construction parameters based on the simulated spectrum, thereby obtaining the sensitivity of each library construction parameter; the weight division module obtains the division weight of each library construction parameter based on the sensitivity of each library construction parameter and the parameter value range of each library construction parameter; the grid division module calculates the number of grid divisions for each library construction parameter based on the division weight of each library construction parameter; the parameter network acquisition module obtains grid points based on the number of grid divisions for each library construction parameter, and each grid point corresponds to a set of geometric parameters to obtain a parameter network; finally, the spectral library optimization module uses the Fourier modal method to simulate the geometric parameters corresponding to each grid point in the parameter network to generate an optimized spectral library.

[0044] Furthermore, the sensitivity analysis module is used to perform sensitivity analysis on the preset library construction parameters based on the simulated spectrum to obtain the sensitivity of each library construction parameter, including:

[0045] The sensitivity of each library construction parameter is calculated using the partial derivative finite difference method, which is expressed as:

[0046]

[0047] Where F_λ represents the simulated spectrum at the λth wavelength, p_i represents the process expected value of the i-th library construction parameter, Δp_i represents the parameter perturbation step, N represents the wavelength index, and paraSense[i] represents the sensitivity of the i-th library construction parameter.

[0048] The above scheme uses the finite difference method to calculate partial derivatives to determine the impact of changes in library construction parameters on optical measurement results. Specifically, the sensitivity of each library construction parameter, paraSense[i], is obtained by simulating the output change of the spectrum, providing an important basis for subsequent grid division. In addition to calculating the maximum value of the output change of the simulated spectrum as described in the above scheme, the sensitivity of each library construction parameter can also be obtained by calculating the average value or other variations. Compared to the existing method of manually setting the sampling interval, the above scheme integrates sensitivity analysis into automatic grid division, achieving dynamic adjustment of sampling density and improving accuracy.

[0049] Furthermore, the weight division module is used to obtain the parameter value range of each library construction parameter, and based on the sensitivity of each library construction parameter and the parameter value range of each library construction parameter, obtain the division weight of each library construction parameter, including:

[0050] According to the database construction parameters, calculate the value range of each database construction parameter paraRange[i], which represents the value range of the i-th database construction parameter;

[0051] Multiply the sensitivity of each library construction parameter by the parameter value range of each library construction parameter to obtain the partition weight of each library construction parameter, which is specifically expressed as: paraWeight[i]=paraSense[i]×paraRange[i];

[0052] Where paraWeight[i] represents the partition weight of the i-th database construction parameter, paraSense[i] represents the sensitivity of the i-th database construction parameter, and paraRange[i] represents the parameter value range of the i-th database construction parameter.

[0053] The above scheme calculates the partitioning weight of each library construction parameter and assigns higher weights to more representative library construction parameter data, so that this data plays a greater role in the subsequent steps, thereby improving the accuracy of the neural network training set.

[0054] Furthermore, the parameter network acquisition module is used to obtain grid points according to the number of grid divisions of each library construction parameter, and each grid point corresponds to a set of geometric parameters, including:

[0055] The corresponding grid points are calculated based on the number of grid divisions for each parameter. Each grid point corresponds to a set of geometric parameters, expressed as:

[0056]

[0057] Where paraMin[i] represents the minimum value of the i-th database construction parameter, Grid[i] represents the number of grid divisions for the i-th database construction parameter, j represents any positive integer less than Grid[i], and m represents the number of database construction parameters.

[0058] The geometric parameters corresponding to each grid point in the above scheme can form a parameter network. Then, based on the parameter network sampling, the spectral library of the geometric parameters corresponding to the grid point is simulated by the Fourier modal method to generate an optimized neural network training set, thereby realizing an automatic optimization scheme based on sensitivity analysis, ensuring that the sampling points are concentrated in the effective area and improving accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A schematic flow chart of a spectral library optimization method based on sensitivity analysis provided in one embodiment of the present invention;

[0060] Figure 2 A semiconductor example is provided to illustrate a method and system for optimizing a spectral library based on sensitivity analysis according to an embodiment of the present invention;

[0061] Figure 3 A schematic diagram of a spectral library optimization system architecture based on sensitivity analysis provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0062] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0063] See Figure 1 This embodiment provides a spectral library optimization method based on sensitivity analysis, comprising the following steps:

[0064] Step S1: obtaining a simulated spectrum;

[0065] Step S2: performing sensitivity analysis on the preset library construction parameters based on the simulated spectra to obtain the sensitivity of each library construction parameter;

[0066] Step S3: Obtain the parameter value range of each library construction parameter, and obtain the partition weight of each library construction parameter based on the sensitivity of each library construction parameter and the parameter value range of each library construction parameter;

[0067] Step S4: Calculate the number of grid divisions for each database construction parameter based on the division weight of each database construction parameter;

[0068] Step S5: Obtain grid points according to the number of grid divisions of each database parameter, each grid point corresponds to a set of geometric parameters, and obtain a parameter network;

[0069] Step S6: The geometric parameters corresponding to each grid point in the parameter network are simulated by the Fourier modal method to generate an optimized spectrum library.

[0070] In this embodiment, by obtaining a simulated spectrum, a sensitivity analysis is performed on the preset library construction parameters, that is, the change of the library construction parameters to the spectral output is calculated to obtain the sensitivity of each library construction parameter; the division weight of each library construction parameter is obtained according to the parameter value range and sensitivity of each library construction parameter, thereby calculating the number of grid divisions for each library construction parameter; grid points are obtained according to the number of grid divisions for each library construction parameter, wherein a set of geometric parameters corresponding to each grid point constitutes a parameter network; finally, the geometric parameters corresponding to each grid point in the parameter network are simulated by the Fourier modal method to generate an optimized spectral library, and the obtained optimized spectral library is used as a neural network for training. The spectrum sampling samples obtained by the above scheme are evenly distributed, with small sample deviation, which greatly improves the accuracy and stability of the neural network model. In addition, the scheme of the present invention is applicable to multiple optical measurement fields such as film thickness and optical geometric dimensions for generating spectral libraries with different parameter spaces, meeting the needs of training and prediction of multiple optical models.

[0071] In one embodiment, performing sensitivity analysis on preset library construction parameters based on simulated spectra to obtain the sensitivity of each library construction parameter includes:

[0072] The sensitivity of each library construction parameter is calculated using the partial derivative finite difference method, which is expressed as:

[0073]

[0074] Where F_λ represents the simulated spectrum at the λth wavelength, p_i represents the process expected value of the i-th library construction parameter, Δp_i represents the parameter perturbation step, N represents the wavelength index, and paraSense[i] represents the sensitivity of the i-th library construction parameter.

[0075] In the specific implementation process of this embodiment, the finite difference method is used to calculate partial derivatives to determine the impact of changes in library construction parameters on optical measurement results. That is, the sensitivity of each library construction parameter, paraSense[i], is obtained by simulating the output change of the spectrum, providing an important basis for subsequent grid division. In addition to calculating the maximum value of the output change of the simulated spectrum in the above scheme, the sensitivity of each library construction parameter can also be obtained by calculating the average value or other variations. Compared to the existing method of manually setting the sampling interval, the above scheme integrates sensitivity analysis into automatic grid division, realizing dynamic adjustment of sampling density and improving accuracy.

[0076] In one embodiment, the step of obtaining a parameter value range of each library construction parameter and obtaining a partitioning weight of each library construction parameter based on the sensitivity of each library construction parameter and the parameter value range of each library construction parameter specifically includes:

[0077] According to the database construction parameters, calculate the value range of each database construction parameter paraRange[i], which represents the value range of the i-th database construction parameter;

[0078] Multiply the sensitivity of each library construction parameter by the parameter value range of each library construction parameter to obtain the partition weight of each library construction parameter, which is specifically expressed as: paraWeight[i]=paraSense[i]×paraRange[i];

[0079] Where paraWeight[i] represents the partition weight of the i-th database construction parameter, paraSense[i] represents the sensitivity of the i-th database construction parameter, and paraRange[i] represents the parameter value range of the i-th database construction parameter.

[0080] During the specific implementation process, this embodiment calculates the division weight of each library construction parameter and assigns a higher weight to more representative library construction parameter data, so that the data plays a greater role in the next steps, thereby improving the accuracy of the neural network training set.

[0081] In one embodiment, the calculation of the number of grid divisions for each database construction parameter based on the division weight of each database construction parameter specifically includes:

[0082] Normalize each library construction parameter;

[0083] The maximum and minimum weight values ​​of the database construction parameter division weights are normalized and expressed as:

[0084]

[0085] Among them, normParaWeight[i] is the partition weight of the i-th database construction parameter after normalization, paraWeightMax represents the maximum value of the partition weights of all database construction parameters, and paraWeightMin represents the minimum value of the partition weights of all database construction parameters.

[0086] This embodiment combines sensitivity analysis with weighted normalization to assign higher weights to highly sensitive library parameters, making them more important in subsequent calculations. This dynamically adjusts the sampling density for subsequent automatic meshing, providing higher accuracy and stability. Furthermore, this solution is particularly well-suited for ellipsometry, a technique commonly used to characterize critical dimensions and optical constants of semiconductor chips.

[0087] In one embodiment, the calculation of the number of grid divisions for each database construction parameter based on the division weight of each database construction parameter specifically includes:

[0088] Calculate the base value base used for preliminary allocation of grid numbers, expressed as:

[0089]

[0090] Where LibCountUser represents the number of preset spectral libraries, and m represents the number of library construction parameters;

[0091] The number of grid divisions for each database construction parameter is calculated based on the normalized division weight and basic value, which can be expressed as follows:

[0092] Grid[i]=normParaWeight[i]×base;

[0093] Among them, Grid[i] represents the number of grid divisions of the i-th database construction parameter.

[0094] This implementation uses the user-specified number of spectral libraries and the number of library construction parameters to determine a base value for initial grid number allocation. The base value then calculates the optimal number of grid divisions for each library construction parameter based on the normalized division weights and the base value. By comprehensively considering the sensitivity, value range, and division weight of each library construction parameter, the automatically calculated optimal number of grid divisions avoids the problem of irrational grid divisions caused by manual settings, effectively improving computational efficiency and accuracy.

[0095] In one embodiment, the method of obtaining grid points according to the number of grid divisions of each library parameter, wherein each grid point corresponds to a set of geometric parameters, and obtaining a parameter network specifically includes:

[0096] The corresponding grid points are calculated based on the number of grid divisions for each parameter. Each grid point corresponds to a set of geometric parameters, expressed as:

[0097]

[0098] Where paraMin[i] represents the minimum value of the i-th database construction parameter, Grid[i] represents the number of grid divisions for the i-th database construction parameter, j represents any positive integer less than Grid[i], and m represents the number of database construction parameters.

[0099] In the specific implementation process of this embodiment, a parameter network can be constructed through the geometric parameters corresponding to each grid point. Then, based on the parameter network sampling, the spectral library of the geometric parameters corresponding to the grid point is simulated by the Fourier modal method to generate an optimized neural network training set, thereby realizing an automatic optimization scheme based on sensitivity analysis, ensuring that the sampling points are concentrated in the effective area, and improving accuracy.

[0100] In order to more clearly illustrate the spectral library optimization method based on sensitivity analysis provided by the above scheme and highlight its technical advantages, this example uses Figure 2 The semiconductor example shown is used for illustration. Figure 2 The specific meanings involved are shown in Table 1 below:

[0101] Table 1 Specific description of semiconductor examples

[0102]

[0103] This embodiment compares the number of grid divisions of the automatically generated (Auto) database construction parameter with the manual setting (SelfDefine) method in the prior art. The specific data are shown in Table 1 below:

[0104] Table 2 Grid allocation table under different methods

[0105]

[0106]

[0107] As shown in Table 2, Auto represents the spectral library optimization method based on sensitivity analysis adopted in this embodiment; SelfDefine represents the manual setting method adopted in the prior art; L1_HT represents the thickness of the first layer of film; G2_Depth represents the height of the trapezoid in the second layer;

[0108] Based on the Auto and SelfDefine datasets described in Table 2, two neural network models were trained under the same conditions. Subsequently, regression tests were performed using simulated spectra (L1_HT: 90, G2_Depth: 40). The results are as follows:

[0109] Table 3 Comparison of results under different methods

[0110]

[0111] Through the above steps, it can be seen from the relevant data results provided in Table 3 that the spectral library optimization system based on sensitivity analysis applied in this embodiment automatically generates a reasonable sampling point distribution in the specified geometric parameter space and dynamically adjusts the sampling density. It can effectively avoid the blindness of human operation and dependence on experience, improve the automation and intelligence level of the sampling process, realize efficient and automatic grid division calculation, provide a scientific basis for parameter allocation in the library generation process, and effectively improve the accuracy of the final result.

[0112] See Figure 3 This embodiment also provides a spectral library optimization system based on sensitivity analysis, including: a measurement result acquisition module for acquiring simulated spectra; a sensitivity analysis module for performing sensitivity analysis on preset library construction parameters based on the simulated spectra to obtain the sensitivity of each library construction parameter; a weight division module for obtaining the parameter value range of each library construction parameter, and obtaining the division weight of each library construction parameter based on the sensitivity of each library construction parameter and the parameter value range of each library construction parameter; a grid division module for calculating the number of grid divisions for each library construction parameter based on the division weight of each library construction parameter; a parameter network acquisition module for obtaining grid points according to the number of grid divisions for each library construction parameter, each grid point corresponding to a set of geometric parameters, to obtain a parameter network; and a spectral library optimization module for simulating the geometric parameters corresponding to each grid point in the parameter network by the Fourier modal method to generate an optimized spectral library.

[0113] The system architecture proposed in this embodiment is simple and easy to implement, and can be used to implement a spectral library optimization method and system based on sensitivity analysis as described above. A simulated spectrum is obtained through a measurement result acquisition module; a sensitivity analysis module performs sensitivity analysis on preset library construction parameters based on the simulated spectrum, thereby obtaining the sensitivity of each library construction parameter; a weight division module is used to obtain the division weight of each library construction parameter based on the sensitivity of each library construction parameter and the parameter value range of each library construction parameter; a grid division module is used to calculate the number of grid divisions for each library construction parameter based on the division weight of each library construction parameter; a parameter network acquisition module is used to obtain grid points based on the number of grid divisions for each library construction parameter, each grid point corresponding to a set of geometric parameters, to obtain a parameter network; finally, a spectral library optimization module is used to simulate the geometric parameters corresponding to each grid point in the parameter network using the Fourier modal method to generate an optimized spectral library.

[0114] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A spectral library optimization method based on sensitivity analysis, characterized in that: include: Get simulated spectra; Perform sensitivity analysis on the preset library construction parameters based on simulated spectra to obtain the sensitivity of each library construction parameter; Obtain the parameter value range of each library construction parameter, and obtain the partition weight of each library construction parameter based on the sensitivity of each library construction parameter and the parameter value range of each library construction parameter; Calculate the number of grid divisions for each database construction parameter based on the division weight of each database construction parameter; Obtain grid points based on the number of grid divisions for each database parameter. Each grid point corresponds to a set of geometric parameters to obtain a parameter network. The geometric parameters corresponding to each grid point in the parameter network are simulated by the Fourier modal method to generate an optimized spectral library.

2. The method for optimizing a spectral library based on sensitivity analysis according to claim 1, wherein: The sensitivity analysis of the preset library construction parameters based on the simulated spectrum to obtain the sensitivity of each library construction parameter includes: The sensitivity of each library construction parameter is calculated using the partial derivative finite difference method, which is expressed as: Where F_λ represents the simulated spectrum at the λth wavelength, p_i represents the process expected value of the i-th library construction parameter, Δp_i represents the parameter perturbation step, N represents the wavelength index, and paraSense[i] represents the sensitivity of the i-th library construction parameter.

3. The method for optimizing a spectral library based on sensitivity analysis according to claim 2, wherein: The step of obtaining the parameter value range of each library construction parameter and obtaining the partitioning weight of each library construction parameter based on the sensitivity of each library construction parameter and the parameter value range of each library construction parameter specifically includes: According to the database construction parameters, calculate the value range of each database construction parameter paraRange[i], which represents the value range of the i-th database construction parameter; Multiply the sensitivity of each library construction parameter by the parameter value range of each library construction parameter to obtain the partition weight of each library construction parameter, which is specifically expressed as: paraWeight[i]=paraSense[i]×paraRange[i]; Where paraWeight[i] represents the partition weight of the i-th database construction parameter, paraSense[i] represents the sensitivity of the i-th database construction parameter, and paraRange[i] represents the parameter value range of the i-th database construction parameter.

4. The method for optimizing a spectral library based on sensitivity analysis according to claim 3, wherein: The calculation of the number of grid divisions for each database construction parameter based on the division weight of each database construction parameter specifically includes: Normalize each library construction parameter; The maximum and minimum weight values ​​of the database construction parameter division weights are normalized and expressed as: Among them, normParaWeight[i] is the partition weight of the i-th database construction parameter after normalization, paraWeightMax represents the maximum value of the partition weights of all database construction parameters, and paraWeightMin represents the minimum value of the partition weights of all database construction parameters.

5. The method for optimizing a spectral library based on sensitivity analysis according to claim 4, wherein: The calculation of the number of grid divisions for each database construction parameter based on the division weight of each database construction parameter specifically includes: Calculate the base value base used for preliminary allocation of grid numbers, expressed as: Where LibCountUser represents the number of preset spectral libraries, and m represents the number of library construction parameters; The number of grid divisions for each database construction parameter is calculated based on the normalized division weight and basic value, which can be expressed as follows: Grid[i]=normParaWeight[i]×base; Among them, Grid[i] represents the number of grid divisions of the i-th database construction parameter.

6. The method for optimizing a spectral library based on sensitivity analysis according to claim 3, wherein: The grid points are obtained according to the number of grid divisions for each database construction parameter, each grid point corresponds to a set of geometric parameters, and the parameter network is obtained, specifically including: The corresponding grid points are calculated based on the number of grid divisions for each parameter. Each grid point corresponds to a set of geometric parameters, expressed as: Where paraMin[i] represents the minimum value of the i-th database construction parameter, Grid[i] represents the number of grid divisions for the i-th database construction parameter, j represents any positive integer less than Grid[i], and m represents the number of database construction parameters.

7. A spectral library optimization system based on sensitivity analysis, characterized in that: include: A measurement result acquisition module is used to obtain a simulated spectrum; Sensitivity analysis module, used to perform sensitivity analysis on preset library construction parameters based on simulated spectra to obtain the sensitivity of each library construction parameter; The weight division module is used to obtain the parameter value range of each library construction parameter, and obtain the division weight of each library construction parameter based on the sensitivity of each library construction parameter and the parameter value range of each library construction parameter; A grid division module is used to calculate the number of grid divisions for each database construction parameter based on the division weight of each database construction parameter; The parameter network acquisition module is used to obtain grid points according to the number of grid divisions of each database parameter. Each grid point corresponds to a set of geometric parameters to obtain the parameter network. The spectral library optimization module is used to simulate the geometric parameters corresponding to each grid point in the parameter network through the Fourier modal method to generate an optimized spectral library.

8. The spectral library optimization system based on sensitivity analysis according to claim 7, characterized in that: The sensitivity analysis module is used to perform sensitivity analysis on the preset library construction parameters based on the simulated spectrum to obtain the sensitivity of each library construction parameter, including: The sensitivity of each library construction parameter is calculated using the partial derivative finite difference method, which is expressed as: Where F_λ represents the simulated spectrum at the λth wavelength, p_i represents the process expected value of the i-th library construction parameter, Δp_i represents the parameter perturbation step, N represents the wavelength index, and paraSense[i] represents the sensitivity of the i-th library construction parameter.

9. The spectral library optimization system based on sensitivity analysis according to claim 7, characterized in that: The weight division module is used to obtain the parameter value range of each library construction parameter, and obtain the division weight of each library construction parameter based on the sensitivity of each library construction parameter and the parameter value range of each library construction parameter, including: According to the database construction parameters, calculate the value range of each database construction parameter paraRange[i], which represents the value range of the i-th database construction parameter; Multiply the sensitivity of each library construction parameter by the parameter value range of each library construction parameter to obtain the partition weight of each library construction parameter, which is specifically expressed as: paraWeight[i]=paraSense[i]×paraRange[i]; Where paraWeight[i] represents the partition weight of the i-th database construction parameter, paraSense[i] represents the sensitivity of the i-th database construction parameter, and paraRange[i] represents the parameter value range of the i-th database construction parameter.

10. The spectral library optimization system based on sensitivity analysis according to claim 7, characterized in that: The parameter network acquisition module is used to obtain grid points according to the number of grid divisions of each database construction parameter. Each grid point corresponds to a set of geometric parameters, including: The corresponding grid points are calculated based on the number of grid divisions for each parameter. Each grid point corresponds to a set of geometric parameters, expressed as: Where paraMin[i] represents the minimum value of the i-th database construction parameter, Grid[i] represents the number of grid divisions for the i-th database construction parameter, j represents any positive integer less than Grid[i], and m represents the number of database construction parameters.

Images