Material formula optimization method, device, electronic device and storage medium

Abstract

The present application belongs to the technical field of optimizing material formula, and discloses a material formula optimization method, device, electronic device and storage medium. The method comprises: using an isolation forest algorithm to remove abnormal data from material spectrum data of each luminescent material to obtain sample spectrum data, obtaining measured spectrum intensity and actual spectrum intensity of each sample spectrum data, and combining the spectrum intensity calculation formula to construct a transmittance calculation model, and using the L‑BFGS algorithm to determine the optimal formula combination of multiple material combinations according to the transmittance calculation model and the functional relationship between the amount of different materials added and the actual spectrum intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the expected coordinates; determining the optimal formula combination of the material combination by using the L‑BFGS algorithm, the transmittance calculation model, the functional relationship between the amount of different materials added and the actual spectrum intensity, the expected coordinates and the CIE coordinates, so as to optimize the material formula and improve the optimization efficiency of the material formula.

Description

Technical Field

[0001] The present application relates to the technical field of optimizing material formulations, and in particular, to a material formulation optimization method, device, electronic device, and storage medium. Background Art

[0002] The optimization and screening of multiple material formulas is the key to the application of new materials. Traditional formula optimization methods often rely on manual trial and error, which has limitations such as long time consumption, high cost, low accuracy, and difficulty in dealing with the problem of multi-factor interaction in complex formula systems.

[0003] With the rapid development of machine learning technology and spectral detection technology, the combination of spectral detection and analysis technology with machine learning algorithms can provide a new solution for the optimization of material formulations. Especially in the high-throughput screening and optimization of material formulations, machine learning algorithms have shown great potential and advantages.

[0004] Therefore, in order to solve the limitations of traditional material formula optimization methods such as long time consumption, high cost, low accuracy, and difficulty in dealing with the technical problems of multi-factor interactions in complex formula systems, a material formula optimization method, device, electronic device and storage medium are urgently needed. Summary of the invention

[0005] The purpose of the present application is to provide a material formula optimization method, device, electronic device and storage medium. Through the L-BFGS algorithm, transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, the optimal formula combination of multiple material combinations is determined to optimize the material formula, solve the limitations of traditional material formula optimization methods such as long time consumption, high cost, low accuracy and difficulty in dealing with the interaction of multiple factors in complex formula systems, and can accurately and quickly optimize the formula, thereby improving the optimization efficiency of the material formula.

[0006] In a first aspect, the present application provides a material formulation optimization method, comprising:

[0007] Obtain material spectrum data of luminescent materials in the database;

[0008] Using an isolation forest algorithm, removing abnormal data from the spectral data of each material, and obtaining sample spectral data of each luminescent material;

[0009] Obtaining the measured spectral intensity and the actual spectral intensity of each sample spectral data, and combining them with a preset spectral intensity calculation formula to construct a transmittance calculation model;

[0010] Using the L-BFGS algorithm, according to the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, the optimal formula combination of multiple material combinations is determined; the material combination is a combination of multiple luminescent materials selected from the database.

[0011] The material formula optimization method provided in the present application can optimize the material formula. Through the L-BFGS algorithm, the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, the optimal formula combination of multiple material combinations is determined to optimize the material formula, so as to solve the limitations of traditional material formula optimization methods such as long time consumption, high cost, low accuracy and difficulty in dealing with the interaction of multiple factors in complex formula systems. The method can accurately and quickly optimize the formula and improve the optimization efficiency of the material formula.

[0012] Optionally, using an isolation forest algorithm to remove abnormal data from the spectral data of each material to obtain sample spectral data of each luminescent material, including:

[0013] Based on the characteristic data of the spectral data points, the spectral data points in the spectral data of each material are divided to obtain an isolated forest partition corresponding to each spectral data point in the spectral data of each material;

[0014] According to the path length of each spectral data point in the corresponding isolated forest partition in the spectral data of each material, combined with a preset abnormal isolation threshold, abnormal data in the spectral data of each material is determined;

[0015] The abnormal data is removed from the spectrum data of each material to obtain the sample spectrum data.

[0016] The material formula optimization method provided in the present application can optimize the material formula. Through the isolation forest algorithm, abnormal data in the material spectral data is eliminated to obtain normal spectral data points. The material formula is optimized based on the normal spectral data points, which is beneficial to improving the optimization efficiency of the material formula.

[0017] Optionally, based on the characteristic data of the spectral data points, the spectral data points in each of the material spectral data are divided to obtain an isolated forest partition corresponding to each spectral data point in each of the material spectral data, including:

[0018] A1, selecting any target feature data from the feature data of the spectral data points, and setting a segmentation threshold corresponding to the target feature data;

[0019] A2, dividing the spectral data points in the spectral data of each material based on the target feature data and the segmentation threshold to form different divided areas;

[0020] A3, repeating steps A1 and A2 until the number of divided areas is greater than a preset division threshold, or the number of spectral data points in all the divided areas is less than or equal to the preset number threshold, to obtain isolated forest partitions corresponding to each spectral data point in the spectral data of each material.

[0021] Optionally, determining the abnormal data in each material spectral data according to the path length of each spectral data point in the corresponding isolation forest partition in the spectral data of each material in combination with a preset abnormal isolation threshold value comprises:

[0022] Calculating the path length of each spectral data point in the spectral data of each material in the corresponding isolated forest partition;

[0023] Determine whether the absolute value of the difference between the path length of any spectral data point in each of the material spectral data and the average path length of other spectral data points is greater than the preset abnormal isolation threshold;

[0024] If yes, then determining that any of the spectral data points is abnormal data;

[0025] If not, it is determined that any of the spectral data points is normal data.

[0026] Optionally, the measured spectral intensity and the actual spectral intensity of each sample spectral data are obtained to construct a transmittance calculation model in combination with a preset spectral intensity calculation formula, including:

[0027] Measuring and obtaining the measured spectral intensity of each sample spectral data;

[0028] The actual spectral intensity of each sample spectral data is measured and obtained by using a quadratic spline function interpolation method;

[0029] A transmittance calculation model is constructed according to the measured spectral intensity, the actual spectral intensity and the preset spectral intensity calculation formula.

[0030] Optionally, the L-BFGS algorithm is used to determine the optimal formula combination of the plurality of material combinations according to the transmittance calculation model and the preset functional relationship between the amount of different materials added and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, including:

[0031] arbitrarily selecting a plurality of the luminescent materials from the database for combination to obtain a plurality of material combinations;

[0032] According to the transmittance calculation model, a least squares function is constructed by combining the CIE coordinates corresponding to the plurality of material combinations, the preset functional relationship between the doping amount of different materials and the actual spectral intensity, and the preset expected coordinates;

[0033] The least squares function is solved by using the L-BFGS algorithm to iterate the CIE coordinates corresponding to each of the material combinations in turn, and when the difference between the iterated CIE coordinates of each of the material combinations and the preset expected coordinates is within a preset error range, the dosage corresponding to the iterated CIE coordinates is determined to be the optimal formula combination of the corresponding material combination, thereby obtaining multiple target formulas.

[0034] The material formula optimization method provided in the present application can optimize the material formula. The L-BFGS algorithm is used to solve the least squares function to iterate the CIE coordinates corresponding to each material combination in turn. When the difference between the iterated CIE coordinates of each material combination and the preset expected coordinates is within a preset error range, the added amount corresponding to the iterative CIE coordinates is determined to be the optimal formula combination of the corresponding material combination. Through the L-BFGS algorithm and the least squares function, the optimal formula combination of each material combination can be quickly calculated, which is conducive to improving the optimization efficiency of the material formula.

[0035] Optionally, the preset error range includes a preset horizontal coordinate error range, a preset vertical coordinate error range and a preset vertical coordinate error range; when the difference between the iterative CIE coordinates of each material combination and the preset expected coordinates is within the preset error range, determining the addition amount corresponding to the iterative CIE coordinates as the optimal formula combination of the corresponding material combination includes:

[0036] When the differences between the CIE coordinates corresponding to the iterations of each material combination and the preset expected coordinates in the horizontal coordinate, vertical coordinate and vertical coordinate are respectively within the preset horizontal coordinate error range, the preset vertical coordinate error range and the preset vertical coordinate error range, the addition amount corresponding to the iterative CIE coordinates is determined to be the optimal formula combination of the corresponding material combination.

[0037] In a second aspect, the present application provides a material formulation optimization device, comprising:

[0038] An acquisition module, used for acquiring material spectrum data of luminescent materials in a database;

[0039] A removal module, used to remove abnormal data from the spectral data of each material using an isolation forest algorithm to obtain sample spectral data of each luminescent material;

[0040] A construction module, used to obtain the measured spectral intensity and the actual spectral intensity of each sample spectral data, and to construct a transmittance calculation model in combination with a preset spectral intensity calculation formula;

[0041] The determination module is used to determine the optimal formula combination of multiple material combinations by using the L-BFGS algorithm, according to the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, to obtain multiple target formulas; the material combination is a combination of multiple luminescent materials selected from the database.

[0042] The material formula optimization device determines the optimal formula combination of multiple material combinations through the L-BFGS algorithm, the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, so as to optimize the material formula, solve the limitations of traditional material formula optimization methods such as long time consumption, high cost, low accuracy and difficulty in dealing with the interaction of multiple factors in a complex formula system, and can accurately and quickly optimize the formula, thereby improving the optimization efficiency of the material formula.

[0043] In a third aspect, the present application provides an electronic device comprising a processor and a memory, wherein the memory stores a computer program executable by the processor, and when the processor executes the computer program, it runs the steps in the material formula optimization method described above.

[0044] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps in the material formulation optimization method described above are executed.

[0045] Beneficial effects: The material formula optimization method, device, electronic device and storage medium provided in the present application determine the optimal formula combination of multiple material combinations through the L-BFGS algorithm, transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, to optimize the material formula, solve the limitations of traditional material formula optimization methods such as long time consumption, high cost, low accuracy and difficulty in dealing with the interaction of multiple factors in complex formula systems, and can accurately and quickly optimize the formula, thereby improving the optimization efficiency of the material formula. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A flow chart of a material formulation optimization method provided in an embodiment of the present application.

[0047] Figure 2A schematic diagram of the structure of a material formulation optimization device provided in an embodiment of the present application.

[0048] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0049] Explanation of reference numerals: 1. Acquisition module; 2. Elimination module; 3. Construction module; 4. Determination module; 301. Processor; 302. Memory; 303. Communication bus. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0051] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0052] Please refer to Figure 1 , Figure 1 A material formula optimization method in some embodiments of the present application is used to optimize the material formula, comprising the steps of:

[0053] Step S101, obtaining material spectrum data of luminescent materials in a database;

[0054] Step S102, using an isolation forest algorithm to remove abnormal data from the spectrum data of each material to obtain sample spectrum data of each luminescent material;

[0055] Step S103, obtaining the measured spectral intensity and the actual spectral intensity of each sample spectral data, and combining them with a preset spectral intensity calculation formula to construct a transmittance calculation model;

[0056] Step S104, using the L-BFGS algorithm, according to the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, determine the optimal formula combination of multiple material combinations to obtain multiple target formulas; the material combination is a combination of multiple luminescent materials selected from the database.

[0057] The material formula optimization method determines the optimal formula combination of multiple material combinations through the L-BFGS algorithm, the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, to optimize the material formula, solve the limitations of traditional material formula optimization methods such as long time consumption, high cost, low accuracy and difficulty in dealing with the interaction of multiple factors in complex formula systems, and can accurately and quickly optimize the formula, thereby improving the optimization efficiency of the material formula.

[0058] Specifically, in step S101, material spectrum data of luminescent materials (such as fluorescent materials, metal materials, semiconductor materials, etc.) in a database are obtained, and material spectrum data of associated luminescent materials are obtained according to actual optimization purpose needs.

[0059] The spectral intensity of the material spectrum data is related to the spectral doping amount and the material volume, so when obtaining the material spectrum data, it is necessary to ensure that the volume of each luminescent material is kept as consistent as possible. The spectral doping amount refers to the amount of material doped when making the spectral data. The material spectrum data consists of multiple spectral data points.

[0060] Specifically, in step S102, an isolation forest algorithm is used to remove abnormal data from the spectral data of each material to obtain sample spectral data of each luminescent material, including:

[0061] Based on the characteristic data of the spectral data points, the spectral data points in the spectral data of each material are divided to obtain the isolated forest partition corresponding to each spectral data point in the spectral data of each material;

[0062] According to the path length of each spectral data point in the spectral data of each material in the corresponding isolated forest partition, combined with a preset abnormal isolation threshold, the abnormal data in the spectral data of each material is determined;

[0063] Abnormal data are removed from the spectral data of each material to obtain sample spectral data.

[0064] Specifically, in step S102, based on the characteristic data of the spectral data points, the spectral data points in the spectral data of each material are divided to obtain an isolated forest partition corresponding to each spectral data point in the spectral data of each material, including:

[0065] A1, select any target feature data from the feature data of the spectral data point, and set a segmentation threshold corresponding to the target feature data;

[0066] A2, based on the target feature data and the segmentation threshold, the spectral data points in the spectral data of each material are divided to form different division areas;

[0067] A3, repeat step A1 and step A2 until the number of divided areas is greater than the preset division threshold, or the number of spectral data points in all divided areas is less than or equal to the preset number threshold, to obtain the isolated forest partition corresponding to each spectral data point in the spectral data of each material.

[0068] In step S102, in the spectral data of the same material, all spectral data points have the same characteristic data (that is, the characteristic data of each spectral data point in the spectral data of the same material are basically consistent). Therefore, it is necessary to eliminate abnormal data in the spectral data of the material (abnormal data occurs due to instrument error, sample contamination or other abnormal conditions) to ensure that the spectral data points in the spectral data of the material are normal spectral data points. Through the isolation forest algorithm, any feature data is selected from the feature data of the spectral data points as the target feature data, such as any feature data selected from the feature data such as wavelength, intensity, absorption peak, emission peak, etc. as the target feature data. On the selected feature data (target feature data), a segmentation threshold is arbitrarily set, and it is determined whether the feature data corresponding to each spectral data point is greater than the segmentation threshold. The spectral data points greater than the segmentation threshold are divided into a region (divided region), and the spectral data points less than or equal to the segmentation threshold are divided into another region. Each region is an isolated tree, and at least one region can be obtained each time the division is performed. The above division steps are repeated, and a feature data is arbitrarily selected as the target feature data multiple times (the target feature data selected each time can be the same) and any segmentation threshold is set (when the same target feature data is repeatedly selected, the selected segmentation threshold cannot be the same), so as to divide the spectral data points of each region formed after the division again, until the number of regions of each spectral data point is greater than the preset segmentation threshold, or the number of spectral data points in all regions is less than or equal to the preset number threshold, and all regions (i.e., isolated trees) are combined to obtain the final isolation forest partition. Therefore, the spectral data points of each material spectral data are divided into corresponding isolated forest partitions, taking the spectral data of each material as the division unit. When dividing the region, when the spectral data points in any region remain unchanged, the region can be determined as an isolated region, and the spectral data points in the region are recorded as "isolated" state. The number of divisions of each spectral data point when forming the region is recorded as the effective number of divisions.

[0069] The effective division times of each spectral data point is the path length of the spectral data point in the corresponding isolated forest partition. The spectral data points in the spectral data of the same material have the same characteristic data, and all normal spectral data points basically maintain the same effective division times, while there are differences between abnormal data and normal spectral data points. The abnormal data has different effective division times due to abnormal characteristic data, and the abnormal data is usually isolated earlier during division.

[0070] Specifically, in step S102, according to the path length of each spectral data point in the spectral data of each material in the corresponding isolation forest partition, combined with a preset abnormal isolation threshold, abnormal data in the spectral data of each material is determined, including:

[0071] Calculate the path length of each spectral data point in the corresponding isolated forest partition in the spectral data of each material;

[0072] Determine whether the absolute value of the difference between the path length of any spectral data point in the spectral data of each material and the average path length of other spectral data points is greater than a preset abnormal isolation threshold;

[0073] If so, any spectral data point is determined to be abnormal data;

[0074] If not, any spectral data point is determined to be normal data.

[0075] It should be noted that the path length refers to how many times a spectral data point needs to be segmented to be isolated, that is, it determines the effective number of segmentations of the spectral data point when forming an isolated forest partition.

[0076] In step S102, the path length of each spectral data point in the corresponding isolation forest partition is calculated. Because the abnormal data has different effective partition times (i.e., different path lengths), the path length of each spectral data point is compared with the average path length of other spectral data points. When the absolute value of the difference between the path length of any spectral data point and the average path length of other spectral data points is greater than a preset abnormal isolation threshold, the spectral data point is determined to be abnormal data; when the absolute value of the difference between the path length of any spectral data point and the average path length of other spectral data points is less than or equal to the preset abnormal isolation threshold, the spectral data point is determined to be normal data. The preset abnormal isolation threshold can be set according to actual needs.

[0077] In step S102, abnormal data is removed from the spectrum data of each material to obtain sample spectrum data of each luminescent material.

[0078] Specifically, in step S103, the measured spectral intensity and the actual spectral intensity of each sample spectral data are obtained, and combined with a preset spectral intensity calculation formula, a transmittance calculation model is constructed, including:

[0079] Measure and obtain the measured spectral intensity of each sample spectral data;

[0080] The actual spectral intensity of each sample spectral data is measured and obtained by using the quadratic spline function interpolation method;

[0081] According to the measured spectral intensity, the actual spectral intensity and the preset spectral intensity calculation formula, a transmittance calculation model is constructed.

[0082] In step S103, the sample spectrum data is directly measured by the prior art to obtain a measured spectrum intensity including actual spectrum intensity, background spectrum and noise.

[0083] The spectrum data in the interval where the background peak of the sample spectrum data is located is masked, that is, the spectrum intensity value of the interval is set to 0 to remove the influence of the background spectrum and noise, and the masked sample spectrum data is interpolated using a quadratic spline function to detect the actual spectrum intensity (the actual spectrum intensity is the spectrum intensity of the spectrum data itself after removing the background spectrum and noise). Among them, the quadratic spline function interpolation method is a prior art and will not be described in detail here.

[0084] The measured spectral intensity and actual spectral intensity of each luminescent material are input into the preset transmittance calculation model to calculate the corresponding spectral transmittance. The transmittance calculation model is constructed through the machine learning algorithm according to the corresponding relationship between the actual spectral intensity and the spectral transmittance. The specific spectral intensity calculation formula is: ;

[0085] in, is the measured spectral intensity, a is the actual spectral intensity, b is the blue light background spectrum, e is the noise, is the spectral transmittance, ,in, is the background peak in the spectral data, is the peak value of the blank sample spectrum. The blank sample spectrum refers to the sample spectrum containing only background information in the absence of any substance to be tested.

[0086] Multiple sets of actual spectral intensities and corresponding spectral transmittances are obtained through the above formula, and multiple sets of actual spectral intensities and corresponding spectral transmittances are used as training data and verification data. Machine learning algorithms such as the LightGBM algorithm, the XGBoost algorithm, and the CatBoos algorithm are used to build and train a machine learning algorithm model to obtain a preset transmittance calculation model. Among them, machine learning algorithms (such as the LightGBM algorithm, the XGBoost algorithm, and the CatBoos algorithm) are prior arts and are not described in detail here.

[0087] Through many experiments, it was found that for the same material of the same volume, the actual spectral intensity is linearly correlated with the common logarithm lgm of the doping amount, while the actual spectral intensity of different materials is linearly correlated with the doping amount in different logarithms lgm. Therefore, the functional relationship a(m) between different materials corresponding to any doping amount and the actual spectral intensity can be constructed.

[0088] Specifically, in step S104, the L-BFGS algorithm is used to determine the optimal formula combination of multiple material combinations according to the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, to obtain multiple target formulas, including:

[0089] arbitrarily selecting multiple luminescent materials from a database for combination to obtain multiple material combinations;

[0090] According to the transmittance calculation model, the least squares function is constructed by combining the CIE coordinates corresponding to the multiple material combinations, the preset functional relationship between the doping amount of different materials and the actual spectral intensity, and the preset expected coordinates;

[0091] The L-BFGS algorithm is used to solve the least squares function to iterate the CIE coordinates corresponding to each material combination in turn. When the difference between the iterative CIE coordinates of each material combination and the preset expected coordinates is within a preset error range, the addition amount corresponding to the iterative CIE coordinates is determined to be the optimal formula combination of the corresponding material combination, and multiple target formulas are obtained.

[0092] In step S104, two or more luminescent materials are randomly selected from the plurality of luminescent materials in the database for combination to obtain a plurality of material combinations, wherein the respective luminescent materials in the material combinations maintain the same volume.

[0093] According to the transmittance calculation model, combined with the CIE coordinates corresponding to multiple material combinations, the preset functional relationship between the dosage of different materials and the actual spectral intensity and the preset expected coordinates, a least squares function is constructed. The least squares function is specifically:

[0094] ;

[0095] ;

[0096] in, is the CIE coordinate conversion function; is the actual spectral intensity of the material combination (i.e., the sum of the actual spectral intensities of each luminescent material in the material combination), and n is the total number of luminescent materials in the material combination; is the spectral transmittance corresponding to the actual spectral intensity of the material combination; is the preset expected coordinate; It is a constrained symbol, indicating that the previous expression needs to meet the following specific conditions; For any value of i; is the amount of the i-th material in the material combination, is the maximum amount of each luminescent material in the material combination, is the volume of the nth luminescent material in the material combination; is the functional relationship between the amount of the i-th material in the material combination and the actual spectral intensity; The transmittance calculation model.

[0097] The L-BFGS algorithm uses the gradient information (actual spectral intensity of the material combination) and curvature information (spectral transmittance corresponding to the actual spectral intensity of the material combination) of the objective function to iteratively update the solution vector (the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates) to solve the optimal solution to the least squares problem. In each iteration, the algorithm constructs a search direction based on the gradient information of the current point and the curvature information stored in the previous iteration, and performs a one-dimensional search along this search direction to determine the next iteration point (i.e., changing different doping amounts), and updates the iteration information (gradient information, curvature information, and solution vector). When a specific condition is met (i.e., the difference between the iterative CIE coordinates and the preset expected coordinates is within the preset error range), the algorithm stops iterating, determines the optimal doping amount of each luminescent material in the material combination, and uses the corresponding optimal doping amount as the optimal formula combination of the material combination. Similarly, the optimal formula combination of different material combinations is calculated, and each optimal formula combination is used as the target formula to obtain multiple target formulas.

[0098] Specifically, in step S104, the preset error range includes a preset horizontal coordinate error range, a preset vertical coordinate error range, and a preset vertical coordinate error range; when the difference between the iterative CIE coordinates of each material combination and the preset expected coordinates is within the preset error range, determining the addition amount corresponding to the iterative CIE coordinates as the optimal formula combination of the corresponding material combination includes:

[0099] When the differences between the CIE coordinates corresponding to the iterations of each material combination and the preset expected coordinates in the horizontal coordinate, vertical coordinate and vertical coordinate are respectively within the preset horizontal coordinate error range, the preset vertical coordinate error range and the preset vertical coordinate error range, the addition amount corresponding to the iterative CIE coordinate is determined to be the optimal formula combination of the corresponding material combination.

[0100] In step S104, the preset expected coordinates are represented as the target color output. When the error between the CIE coordinates and the expected coordinates is smaller, the CIE coordinates can present the target color output corresponding to the expected coordinates. Wherein, the preset error range includes a preset horizontal coordinate error range, a preset vertical coordinate error range and a preset vertical coordinate error range, and the corresponding error range can be set according to actual needs. The corresponding expected coordinates can be obtained from the preset coordinate-color relationship table according to actual needs for setting. When the difference between the CIE coordinates and the expected coordinates in the horizontal coordinate, the vertical coordinate and the vertical coordinate are respectively within the preset horizontal coordinate error range, the preset vertical coordinate error range and the preset vertical coordinate error range, the CIE coordinates can present the target color output corresponding to the expected coordinates. For example, assuming that the output color of the expected coordinates is yellow, when the errors between the CIE coordinates and the expected coordinates in the horizontal coordinate, the vertical coordinate and the vertical coordinate are 0.0001, 0.0002 and 0.0001 respectively, the CIE coordinates can present yellow.

[0101] For example, the material formula optimization method of the present application is used to optimize the formula of the fluorescent material. Spectral data of 27 fluorescent materials in different volumes and in different doping ranges are obtained. The abnormal spectrum in each spectral data is eliminated using the isolation forest algorithm, the average value of all characteristic data of the remaining spectral data points in each spectral data is calculated, and the average value of each characteristic data is set to the value of the corresponding characteristic data of the remaining spectral data points to obtain the sample spectral data. The average time for each iteration is 0.1 seconds. Multiple formulas are generated by the L-BFGS algorithm within a given doping range. The formula content includes the type and specific doping amount of the fluorescent material combination, and the expected coordinates, the volume of the fluorescent material and the number of types of the fluorescent material combination can be adjusted according to actual needs. As shown in the following table, the table lists the optimal formula combination of some fluorescent materials, where CIEx=0.4, CIEy=0.4, CIEz=0.2 outputs yellow light, and CIEx=0.3, CIEy=0.5, CIEz=0.2 outputs green light. The optimal formula combination of some fluorescent materials is specifically shown in the following table:

[0102]

[0103] It can be seen from the above table that the material formula optimization method of the present application can quickly optimize the formula and accurately output the desired material formula combination.

[0104] As can be seen from the above, the material formula optimization method obtains the material spectrum data of the luminescent material in the database, uses the isolation forest algorithm to eliminate abnormal data from the spectrum data of each material, obtains the sample spectrum data of each luminescent material, obtains the measured spectrum intensity and the actual spectrum intensity of each sample spectrum data, and uses them to combine with the preset spectrum intensity calculation formula to construct a transmittance calculation model. The L-BFGS algorithm is used to determine the optimal formula combination of multiple material combinations according to the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectrum intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates. Combined, multiple target formulas are obtained, and the material combination is a combination of multiple luminescent materials selected from the database; thus, through the L-BFGS algorithm, the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, the optimal formula combination of the multiple material combinations is determined to optimize the material formula, solve the limitations of traditional material formula optimization methods such as long time consumption, high cost, low accuracy, and difficulty in dealing with the interaction of multiple factors in a complex formula system, and can accurately and quickly optimize the formula, thereby improving the optimization efficiency of the material formula.

[0105] refer to Figure 2 The present application provides a material formula optimization device for optimizing a material formula, comprising:

[0106] Acquisition module 1, used to acquire material spectrum data of luminescent materials in a database;

[0107] Elimination module 2, used to use isolation forest algorithm to eliminate abnormal data from the spectrum data of each material, and obtain sample spectrum data of each luminescent material;

[0108] Construction module 3 is used to obtain the measured spectral intensity and the actual spectral intensity of each sample spectral data, and to construct a transmittance calculation model in combination with a preset spectral intensity calculation formula;

[0109] Determination module 4 is used to determine the optimal formula combination of multiple material combinations by using the L-BFGS algorithm, according to the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, to obtain multiple target formulas; the material combination is a combination of multiple luminescent materials selected from the database.

[0110] The material formula optimization device determines the optimal formula combination of multiple material combinations through the L-BFGS algorithm, the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, so as to optimize the material formula, solve the limitations of traditional material formula optimization methods such as long time consumption, high cost, low accuracy and difficulty in dealing with the interaction of multiple factors in a complex formula system, and can accurately and quickly optimize the formula, thereby improving the optimization efficiency of the material formula.

[0111] Specifically, when the acquisition module 1 is executed, it acquires the material spectrum data of the luminescent materials (such as fluorescent materials, metal materials, semiconductor materials, etc.) in the database, and acquires the material spectrum data of the associated luminescent materials according to the actual optimization purpose.

[0112] The spectral intensity of the material spectrum data is related to the spectral doping amount and the material volume, so when obtaining the material spectrum data, it is necessary to ensure that the volume of each luminescent material is kept as consistent as possible. The spectral doping amount refers to the amount of material doped when making the spectral data. The material spectrum data consists of multiple spectral data points.

[0113] Specifically, when the elimination module 2 uses the isolation forest algorithm to eliminate abnormal data from the spectrum data of each material and obtains the sample spectrum data of each luminescent material, it executes:

[0114] Based on the characteristic data of the spectral data points, the spectral data points in the spectral data of each material are divided to obtain the isolated forest partition corresponding to each spectral data point in the spectral data of each material;

[0115] According to the path length of each spectral data point in the spectral data of each material in the corresponding isolated forest partition, combined with a preset abnormal isolation threshold, the abnormal data in the spectral data of each material is determined;

[0116] Abnormal data are removed from the spectral data of each material to obtain sample spectral data.

[0117] Specifically, when the elimination module 2 divides the spectral data points in the spectral data of each material based on the characteristic data of the spectral data points to obtain the isolated forest partitions corresponding to each spectral data point in the spectral data of the material, the elimination module 2 executes:

[0118] A1, select any target feature data from the feature data of the spectral data point, and set a segmentation threshold corresponding to the target feature data;

[0119] A2, based on the target feature data and the segmentation threshold, the spectral data points in the spectral data of each material are divided to form different division areas;

[0120] A3, repeat step A1 and step A2 until the number of divided areas is greater than the preset division threshold, or the number of spectral data points in all divided areas is less than or equal to the preset number threshold, to obtain the isolated forest partition corresponding to each spectral data point in the spectral data of each material.

[0121] When the elimination module 2 is executed, in the spectral data of the same material, all spectral data points have the same characteristic data (that is, the characteristic data of each spectral data point in the spectral data of the same material are basically consistent). Therefore, it is necessary to eliminate abnormal data in the spectral data of the material (abnormal data caused by instrument errors, sample contamination or other abnormal conditions) to ensure that the spectral data points in the spectral data of the material are normal spectral data points. Through the isolation forest algorithm, any feature data is selected from the feature data of the spectral data points as the target feature data, such as any feature data selected from the feature data such as wavelength, intensity, absorption peak, emission peak, etc. as the target feature data. On the selected feature data (target feature data), a segmentation threshold is arbitrarily set, and it is determined whether the feature data corresponding to each spectral data point is greater than the segmentation threshold. The spectral data points greater than the segmentation threshold are divided into a region (divided region), and the spectral data points less than or equal to the segmentation threshold are divided into another region. Each region is an isolated tree, and at least one region can be obtained each time the division is performed. The above division steps are repeated, and a feature data is arbitrarily selected as the target feature data multiple times (the target feature data selected each time can be the same) and any segmentation threshold is set (when the same target feature data is repeatedly selected, the selected segmentation threshold cannot be the same), so as to divide the spectral data points of each region formed after the division again, until the number of regions of each spectral data point is greater than the preset segmentation threshold, or the number of spectral data points in all regions is less than or equal to the preset number threshold, and all regions (i.e., isolated trees) are combined to obtain the final isolation forest partition. Therefore, the spectral data points of each material spectral data are divided into corresponding isolated forest partitions, taking the spectral data of each material as the division unit. When dividing the region, when the spectral data points in any region remain unchanged, the region can be determined as an isolated region, and the spectral data points in the region are recorded as "isolated" state. The number of divisions of each spectral data point when forming the region is recorded as the effective number of divisions.

[0122] The effective division times of each spectral data point is the path length of the spectral data point in the corresponding isolated forest partition. The spectral data points in the spectral data of the same material have the same characteristic data, and all normal spectral data points basically maintain the same effective division times, while there are differences between abnormal data and normal spectral data points. The abnormal data has different effective division times due to abnormal characteristic data, and the abnormal data is usually isolated earlier during division.

[0123] Specifically, when determining the abnormal data in the spectral data of each material according to the path length of each spectral data point in the corresponding isolated forest partition in the spectral data of each material and combining the preset abnormal isolation threshold, the elimination module 2 executes:

[0124] Calculate the path length of each spectral data point in the corresponding isolated forest partition in the spectral data of each material;

[0125] Determine whether the absolute value of the difference between the path length of any spectral data point in the spectral data of each material and the average path length of other spectral data points is greater than a preset abnormal isolation threshold;

[0126] If so, any spectral data point is determined to be abnormal data;

[0127] If not, any spectral data point is determined to be normal data.

[0128] It should be noted that the path length refers to how many times a spectral data point needs to be segmented to be isolated, that is, it determines the effective number of segmentations of the spectral data point when forming an isolated forest partition.

[0129] When the elimination module 2 is executed, the path length of each spectral data point in the corresponding isolated forest partition is calculated. Because the abnormal data has different effective partition times (i.e., different path lengths), the path length of each spectral data point is compared with the average path length of other spectral data points. When the absolute value of the difference between the path length of any spectral data point and the average path length of other spectral data points is greater than the preset abnormal isolation threshold, the any spectral data point is determined to be abnormal data; when the absolute value of the difference between the path length of any spectral data point and the average path length of other spectral data points is less than or equal to the preset abnormal isolation threshold, the any spectral data point is determined to be normal data. The preset abnormal isolation threshold can be set according to actual needs.

[0130] When the elimination module 2 is executed, abnormal data is eliminated from the spectrum data of each material to obtain sample spectrum data of each luminescent material.

[0131] Specifically, when the construction module 3 obtains the measured spectral intensity and the actual spectral intensity of each sample spectral data to construct a transmittance calculation model in combination with a preset spectral intensity calculation formula, it executes:

[0132] Measure and obtain the measured spectral intensity of each sample spectral data;

[0133] The actual spectral intensity of each sample spectral data is measured and obtained by using the quadratic spline function interpolation method;

[0134] According to the measured spectral intensity, the actual spectral intensity and the preset spectral intensity calculation formula, a transmittance calculation model is constructed.

[0135] When the construction module 3 is executed, the sample spectrum data is directly measured by the prior art to obtain the measured spectrum intensity including the actual spectrum intensity, the background spectrum and the noise.

[0136] The spectrum data in the interval where the background peak of the sample spectrum data is located is masked, that is, the spectrum intensity value of the interval is set to 0 to remove the influence of the background spectrum and noise, and the masked sample spectrum data is interpolated using a quadratic spline function to detect the actual spectrum intensity (the actual spectrum intensity is the spectrum intensity of the spectrum data itself after removing the background spectrum and noise). Among them, the quadratic spline function interpolation method is a prior art and will not be described in detail here.

[0137] The measured spectral intensity and actual spectral intensity of each luminescent material are input into the preset transmittance calculation model to calculate the corresponding spectral transmittance. The transmittance calculation model is constructed through the machine learning algorithm according to the corresponding relationship between the actual spectral intensity and the spectral transmittance. The specific spectral intensity calculation formula is:

[0138] ;

[0139] in, is the measured spectral intensity, a is the actual spectral intensity, b is the blue light background spectrum, e is the noise, is the spectral transmittance, ,in, is the background peak in the spectral data, is the peak value of the blank sample spectrum. The blank sample spectrum refers to the sample spectrum containing only background information in the absence of any substance to be tested.

[0140] Multiple sets of actual spectral intensities and corresponding spectral transmittances are obtained through the above formula, and multiple sets of actual spectral intensities and corresponding spectral transmittances are used as training data and verification data. Machine learning algorithms such as the LightGBM algorithm, the XGBoost algorithm, and the CatBoos algorithm are used to build and train a machine learning algorithm model to obtain a preset transmittance calculation model. Among them, machine learning algorithms (such as the LightGBM algorithm, the XGBoost algorithm, and the CatBoos algorithm) are prior arts and are not described in detail here.

[0141] Through many experiments, it was found that for the same material of the same volume, the actual spectral intensity is linearly correlated with the common logarithm lgm of the doping amount, while the actual spectral intensity of different materials is linearly correlated with the doping amount in different logarithms lgm. Therefore, the functional relationship a(m) between different materials corresponding to any doping amount and the actual spectral intensity can be constructed.

[0142] Specifically, the determination module 4 uses the L-BFGS algorithm to determine the optimal formula combination of multiple material combinations according to the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, and when multiple target formulas are obtained, executes:

[0143] arbitrarily selecting multiple luminescent materials from a database for combination to obtain multiple material combinations;

[0144] According to the transmittance calculation model, the least squares function is constructed by combining the CIE coordinates corresponding to the multiple material combinations, the preset functional relationship between the doping amount of different materials and the actual spectral intensity, and the preset expected coordinates;

[0145] The L-BFGS algorithm is used to solve the least squares function to iterate the CIE coordinates corresponding to each material combination in turn. When the difference between the iterative CIE coordinates of each material combination and the preset expected coordinates is within a preset error range, the addition amount corresponding to the iterative CIE coordinates is determined to be the optimal formula combination of the corresponding material combination, and multiple target formulas are obtained.

[0146] When the determination module 4 is executed, two or more luminescent materials are randomly selected from the multiple luminescent materials in the database for combination to obtain multiple material combinations, wherein each luminescent material in the material combination maintains the same volume.

[0147] According to the transmittance calculation model, combined with the CIE coordinates corresponding to multiple material combinations, the preset functional relationship between the dosage of different materials and the actual spectral intensity and the preset expected coordinates, a least squares function is constructed. The least squares function is specifically:

[0148] ;

[0149] ;

[0150] in, is the CIE coordinate conversion function; is the actual spectral intensity of the material combination (i.e., the sum of the actual spectral intensities of each luminescent material in the material combination), and n is the total number of luminescent materials in the material combination; is the spectral transmittance corresponding to the actual spectral intensity of the material combination; is the preset expected coordinate; It is a constrained symbol, indicating that the previous expression needs to meet the following specific conditions; For any value of i; is the amount of the i-th material in the material combination, is the maximum amount of each luminescent material in the material combination, is the volume of the nth luminescent material in the material combination; is the functional relationship between the amount of the i-th material in the material combination and the actual spectral intensity; The transmittance calculation model.

[0151] The L-BFGS algorithm uses the gradient information (actual spectral intensity of the material combination) and curvature information (spectral transmittance corresponding to the actual spectral intensity of the material combination) of the objective function to iteratively update the solution vector (the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates) to solve the optimal solution to the least squares problem. In each iteration, the algorithm constructs a search direction based on the gradient information of the current point and the curvature information stored in the previous iteration, and performs a one-dimensional search along this search direction to determine the next iteration point (i.e., changing different doping amounts), and updates the iteration information (gradient information, curvature information, and solution vector). When a specific condition is met (i.e., the difference between the iterative CIE coordinates and the preset expected coordinates is within the preset error range), the algorithm stops iterating, determines the optimal doping amount of each luminescent material in the material combination, and uses the corresponding optimal doping amount as the optimal formula combination of the material combination. Similarly, the optimal formula combination of different material combinations is calculated, and each optimal formula combination is used as the target formula to obtain multiple target formulas.

[0152] Specifically, the preset error range includes a preset horizontal coordinate error range, a preset vertical coordinate error range and a preset vertical coordinate error range; when the difference between the iterative CIE coordinates of each material combination and the preset expected coordinates is within the preset error range, the determination module 4 determines that the addition amount corresponding to the iterative CIE coordinates is the optimal formula combination of the corresponding material combination, and executes:

[0153] When the differences between the CIE coordinates corresponding to the iterations of each material combination and the preset expected coordinates in the horizontal coordinate, vertical coordinate and vertical coordinate are respectively within the preset horizontal coordinate error range, the preset vertical coordinate error range and the preset vertical coordinate error range, the addition amount corresponding to the iterative CIE coordinate is determined to be the optimal formula combination of the corresponding material combination.

[0154] When the determination module 4 is executed, the preset expected coordinates are expressed as the target color output. When the error between the CIE coordinates and the expected coordinates is smaller, the CIE coordinates can present the target color output corresponding to the expected coordinates. Among them, the preset error range includes a preset horizontal coordinate error range, a preset vertical coordinate error range and a preset vertical coordinate error range. The corresponding error range can be set according to actual needs. The corresponding expected coordinates can be obtained from the preset coordinate-color relationship table according to actual needs for setting. When the difference between the CIE coordinates and the expected coordinates in the horizontal coordinate, the vertical coordinate and the vertical coordinate are respectively within the preset horizontal coordinate error range, the preset vertical coordinate error range and the preset vertical coordinate error range, the CIE coordinates can present the target color output corresponding to the expected coordinates. For example, assuming that the output color of the expected coordinates is yellow, when the errors between the CIE coordinates and the expected coordinates in the horizontal coordinate, the vertical coordinate and the vertical coordinate are 0.0001, 0.0002 and 0.0001 respectively, the CIE coordinates can present yellow.

[0155] For example, the material formula optimization device of the present application is used to optimize the formula of the fluorescent material. Spectral data of 27 fluorescent materials in different volumes and in different doping ranges are obtained. The isolated forest algorithm is used to eliminate abnormal spectra in each spectral data, calculate the average value of all characteristic data of the remaining spectral data points in each spectral data, and set the average value of each characteristic data to the value of the corresponding characteristic data of the remaining spectral data points to obtain the sample spectral data. The average time for each iteration is 0.1 seconds. Multiple formulas are generated by the L-BFGS algorithm within a given doping range. The formula content includes the type and specific doping amount of the fluorescent material combination, and the expected coordinates, the volume of the fluorescent material and the number of types of the fluorescent material combination can be adjusted according to actual needs. As shown in the following table, the table lists the optimal formula combination of some fluorescent materials, where CIEx=0.4, CIEy=0.4, CIEz=0.2 outputs yellow light, and CIEx=0.3, CIEy=0.5, CIEz=0.2 outputs green light. The optimal formula combination of some fluorescent materials is specifically shown in the following table:

[0156]

[0157] It can be seen from the above table that the material formula optimization method of the present application can quickly optimize the formula and accurately output the desired material formula combination.

[0158] As can be seen from the above, the material formula optimization device obtains the material spectrum data of the luminescent material in the database, uses the isolation forest algorithm to eliminate abnormal data from the spectrum data of each material, obtains the sample spectrum data of each luminescent material, obtains the measured spectrum intensity and the actual spectrum intensity of each sample spectrum data, and uses them to combine with the preset spectrum intensity calculation formula to construct a transmittance calculation model. The L-BFGS algorithm is used to determine the optimal formula combination of multiple material combinations according to the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectrum intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates. Combined, multiple target formulas are obtained, and the material combination is a combination of multiple luminescent materials selected from the database; thus, through the L-BFGS algorithm, the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, the optimal formula combination of the multiple material combinations is determined to optimize the material formula, solve the limitations of traditional material formula optimization methods such as long time consumption, high cost, low accuracy, and difficulty in dealing with the interaction of multiple factors in a complex formula system, and can accurately and quickly optimize the formula, thereby improving the optimization efficiency of the material formula.

[0159] Please refer to Figure 3 , Figure 3 The present invention provides a structural schematic diagram of an electronic device provided in an embodiment of the present application. The present application provides an electronic device, including: a processor 301 and a memory 302. The processor 301 and the memory 302 are interconnected and communicate with each other through a communication bus 303 and / or other forms of connection mechanisms (not shown). The memory 302 stores a computer program executable by the processor 301. When the electronic device is running, the processor 301 executes the computer program to execute the material formula optimization method in any optional implementation of the above embodiment to achieve the following functions: obtaining material spectrum data of luminescent materials in a database, using an isolation forest algorithm to remove abnormal data from each material spectrum data, obtaining sample spectrum data of each luminescent material, obtaining measured spectrum intensity and actual spectrum intensity of each sample spectrum data, and combining them with a preset spectrum intensity calculation formula to construct a transmittance calculation model, using the L-BFGS algorithm, according to the transmittance calculation model and the preset functional relationship between the amount of different materials added and the actual spectrum intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, determining the optimal formula combination of multiple material combinations, and obtaining multiple target formulas, wherein the material combination is a combination of multiple luminescent materials selected from the database.

[0160] The embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the material formula optimization method in any optional implementation of the above embodiment is executed to achieve the following functions: obtaining material spectrum data of luminescent materials in a database, using an isolation forest algorithm to eliminate abnormal data from each material spectrum data, obtaining sample spectrum data of each luminescent material, obtaining measured spectrum intensity and actual spectrum intensity of each sample spectrum data, and combining them with a preset spectrum intensity calculation formula to construct a transmittance calculation model, using the L-BFGS algorithm, according to the transmittance calculation model and the preset functional relationship between the amount of different materials added and the actual spectrum intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, determine the optimal formula combination of multiple material combinations, and obtain multiple target formulas, where the material combination is a combination of multiple luminescent materials selected from the database. Among them, the storage medium 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, magnetic disk or optical disk.

[0161] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0162] In addition, the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0163] Furthermore, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0164] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0165] The above description is only an embodiment of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A material formula optimization method, used to optimize a material formula, characterized in that: Includes steps: Obtain material spectrum data of luminescent materials in the database; Using an isolation forest algorithm, removing abnormal data from the spectral data of each material, and obtaining sample spectral data of each luminescent material; Obtaining the measured spectral intensity and the actual spectral intensity of each of the sample spectral data, and combining them with a preset spectral intensity calculation formula and a machine learning algorithm to construct a transmittance calculation model according to the corresponding relationship between the actual spectral intensity and the spectral transmittance; Using the L-BFGS algorithm, according to the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, the optimal formula combination of multiple material combinations is determined; the material combination is a combination of multiple luminescent materials selected from the database.

2. The material formula optimization method according to claim 1, characterized in that: Using the isolation forest algorithm, abnormal data is removed from the spectral data of each material to obtain sample spectral data of each luminescent material, including: Based on the characteristic data of the spectral data points, the spectral data points in the spectral data of each material are divided to obtain an isolated forest partition corresponding to each spectral data point in the spectral data of each material; According to the path length of each spectral data point in the corresponding isolated forest partition in the spectral data of each material, combined with a preset abnormal isolation threshold, abnormal data in the spectral data of each material is determined; The abnormal data is removed from the spectrum data of each material to obtain the sample spectrum data.

3. The material formula optimization method according to claim 2, characterized in that: Based on the characteristic data of the spectral data points, the spectral data points in the spectral data of each material are divided to obtain isolated forest partitions corresponding to each spectral data point in the spectral data of each material, including: A1, selecting any target feature data from the feature data of the spectral data points, and setting a segmentation threshold corresponding to the target feature data; A2, based on the target feature data and the segmentation threshold, dividing the spectral data points in the spectral data of each material to form different divided areas; A3, repeating steps A1 and A2 until the number of divided areas is greater than a preset division threshold, or the number of spectral data points in all the divided areas is less than or equal to the preset number threshold, to obtain isolated forest partitions corresponding to each spectral data point in the spectral data of each material.

4. The material formulation optimization method according to claim 2, characterized in that: According to the path length of each spectral data point in the corresponding isolated forest partition in the spectral data of each material, combined with a preset abnormal isolation threshold, abnormal data in the spectral data of each material is determined, including: Calculating the path length of each spectral data point in the spectral data of each material in the corresponding isolated forest partition; Determine whether the absolute value of the difference between the path length of any spectral data point in each of the material spectral data and the average path length of other spectral data points is greater than the preset abnormal isolation threshold; If yes, then determining that any of the spectral data points is abnormal data; If not, it is determined that any of the spectral data points is normal data.

5. The material formulation optimization method according to claim 1, characterized in that: The measured spectral intensity and the actual spectral intensity of each sample spectral data are obtained, and a transmittance calculation model is constructed based on the corresponding relationship between the actual spectral intensity and the spectral transmittance in combination with a preset spectral intensity calculation formula and a machine learning algorithm, including: Measuring and obtaining the measured spectral intensity of each sample spectral data; The actual spectral intensity of each sample spectral data is measured and obtained by using a quadratic spline function interpolation method; According to the measured spectral intensity, the actual spectral intensity and the preset spectral intensity calculation formula, a transmittance calculation model is constructed through the machine learning algorithm and the corresponding relationship between the actual spectral intensity and the spectral transmittance.

6. The material formulation optimization method according to claim 1, characterized in that: Using the L-BFGS algorithm, according to the transmittance calculation model and the preset functional relationship between the amount of different materials added and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, the optimal formula combination of the plurality of material combinations is determined, including: arbitrarily selecting a plurality of the luminescent materials from the database for combination to obtain a plurality of material combinations; According to the transmittance calculation model, a least squares function is constructed by combining the CIE coordinates corresponding to the plurality of material combinations, the preset functional relationship between the doping amount of different materials and the actual spectral intensity, and the preset expected coordinates; The least squares function is solved by using the L-BFGS algorithm to iterate the CIE coordinates corresponding to each of the material combinations in turn, and when the difference between the iterated CIE coordinates of each of the material combinations and the preset expected coordinates is within a preset error range, the dosage corresponding to the iterated CIE coordinates is determined to be the optimal formula combination of the corresponding material combination, thereby obtaining multiple target formulas.

7. The material formula optimization method according to claim 6, characterized in that: The preset error range includes a preset horizontal coordinate error range, a preset vertical coordinate error range and a preset vertical coordinate error range; When the difference between the iterative CIE coordinates of each material combination and the preset expected coordinates is within a preset error range, determining the addition amount corresponding to the iterative CIE coordinates as the optimal formula combination of the corresponding material combination includes: When the differences between the CIE coordinates corresponding to the iterations of each material combination and the preset expected coordinates in the horizontal coordinate, vertical coordinate and vertical coordinate are respectively within the preset horizontal coordinate error range, the preset vertical coordinate error range and the preset vertical coordinate error range, the addition amount corresponding to the iterative CIE coordinates is determined to be the optimal formula combination of the corresponding material combination.

8. A material formula optimization device, used for optimizing material formula, characterized in that: include: An acquisition module, used for acquiring material spectrum data of luminescent materials in a database; A removal module, used to remove abnormal data from the spectral data of each material using an isolation forest algorithm to obtain sample spectral data of each luminescent material; A construction module is used to obtain the measured spectral intensity and the actual spectral intensity of each sample spectral data, and to construct a transmittance calculation model according to the corresponding relationship between the actual spectral intensity and the spectral transmittance in combination with a preset spectral intensity calculation formula and a machine learning algorithm; The determination module is used to determine the optimal formula combination of multiple material combinations by using the L-BFGS algorithm, according to the transmittance calculation model and the preset functional relationship between the dosage of different materials and the actual spectral intensity, combined with the difference between the CIE coordinates corresponding to the material combination and the preset expected coordinates, to obtain multiple target formulas; the material combination is a combination of multiple luminescent materials selected from the database.

9. An electronic device, characterized in that: It comprises a processor and a memory, wherein the memory stores a computer program executable by the processor, and when the processor executes the computer program, it runs the steps in the material formula optimization method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the material formulation optimization method according to any one of claims 1 to 7 are executed.

Images