An adaptive test optimization method and system for element replacement

Abstract

The application discloses an element replacement adaptive test optimization method and system, and relates to the technical field of element test optimization. The method comprises the following steps: extracting performance data and element types of scarce elements in target element replacement scene information, performing first-order multidimensional screening, and obtaining a verified rich element combination set; performing virtual characterization reverse fine-tuning, and obtaining a verified fine-tuning rich element combination set; obtaining a verified fine-tuning rich element combination sample set and a sputtering experiment result set; performing test evaluation according to a preset performance test type, and determining a target fine-tuning rich element combination. The application solves the technical problems of low test efficiency, high research and development loss, difficult performance guarantee and insufficient differentiation of rich element replacement scarce element test in the prior art, achieves the replacement test optimization of rich elements on scarce elements, improves the research and development efficiency, reduces the research and development loss, and guarantees the performance and differentiation potential of the rich element combination.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of element test optimization, and in particular to an adaptive test optimization method and system for element replacement. BACKGROUND

[0002] In the field of high-end coated material target material manufacturing, the performance of the target material directly determines the quality of the coated product, and the traditional target material production relies on scarce elements which have problems such as few resources, high cost, and unstable supply. The replacement of rich elements for scarce elements has become a key direction for the development of the industry. However, the existing test method relies on experience to screen rich elements, which is easy to miss suitable combinations, lacks double control of the performance and differentiation of rich element combinations, and the virtual prediction is disconnected from the experiment, resulting in low test efficiency, high research and development loss, and serious constraints on industrialization.

[0003] The existing technology has the technical problems of low test efficiency, high research and development loss, and difficult guarantee of rich element combination performance and insufficient differentiation when replacing scarce elements with rich elements. SUMMARY

[0004] The present application provides an adaptive test optimization method and system for element replacement, which is used to solve the technical problems of low test efficiency, high research and development loss, and difficult guarantee of rich element combination performance and insufficient differentiation when replacing scarce elements with rich elements in the prior art.

[0005] In view of the above problems, the present application provides an adaptive test optimization method and system for element replacement.

[0006] In a first aspect of the present application, an adaptive test optimization method for element replacement is provided, which comprises:

[0007] Extracting the performance data and type of the scarce element in the target element replacement scenario information, and performing one-dimensional multi-dimensional screening in the rich element resource library to obtain a set of verified rich element combinations; based on double targets, the set of verified rich element combinations is virtually characterized and inversely adjusted to obtain a set of verified and adjusted rich element combinations, wherein the double targets include that the similarity of the set of verified rich element combinations to the performance of the scarce element meets a preset performance requirement and the distribution concentration between the virtual characterization of the set of verified rich element combinations is less than a preset distribution concentration requirement; traversing the set of verified and adjusted rich element combinations to perform sample preparation and sputtering experiments to obtain a set of verified and adjusted rich element combination samples and a set of sputtering experiment results; when the set of sputtering experiment results is passed, the set of verified and adjusted rich element combination samples is tested and evaluated according to a preset performance test type, and a target adjusted rich element combination is determined according to the test evaluation result.

[0008] In a second aspect of the present application, an adaptive test optimization system for element replacement is provided, which comprises:

[0009] The rich element combination set acquisition module is configured to extract the scarce element performance data and the scarce element type in the target element replacement scene information, and perform first-order multi-dimensional screening in the rich element resource library to obtain a to-be-verified rich element combination set. The reverse fine-tuning module is configured to perform virtual representation reverse fine-tuning on the to-be-verified rich element combination set based on double targets to obtain a to-be-verified fine-tuned rich element combination set. The double targets include that the to-be-verified rich element combination satisfies a preset performance requirement in terms of performance similarity with the scarce element, and the distribution concentration between the virtual representation of the to-be-verified rich element combination set is less than a preset distribution concentration requirement. The sputtering experiment module is configured to traverse the to-be-verified fine-tuned rich element combination set to perform sample preparation and sputtering experiment, and obtain a to-be-verified fine-tuned rich element combination sample set and a sputtering experiment result set. The test evaluation module is configured to, when the sputtering experiment result set passes the experiment, perform test evaluation on the to-be-verified fine-tuned rich element combination sample set according to a preset performance test type, and determine a target fine-tuned rich element combination according to a test evaluation result.

[0010] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0011] The rich element combination set acquisition module is configured to extract the scarce element performance data and the scarce element type in the target element replacement scene information, and perform first-order multi-dimensional screening in the rich element resource library to obtain a to-be-verified rich element combination set. The reverse fine-tuning module is configured to perform virtual representation reverse fine-tuning on the to-be-verified rich element combination set based on double targets to obtain a to-be-verified fine-tuned rich element combination set. The double targets include that the to-be-verified rich element combination satisfies a preset performance requirement in terms of performance similarity with the scarce element, and the distribution concentration between the virtual representation of the to-be-verified rich element combination set is less than a preset distribution concentration requirement. The sputtering experiment module is configured to traverse the to-be-verified fine-tuned rich element combination set to perform sample preparation and sputtering experiment, and obtain a to-be-verified fine-tuned rich element combination sample set and a sputtering experiment result set. The test evaluation module is configured to, when the sputtering experiment result set passes the experiment, perform test evaluation on the to-be-verified fine-tuned rich element combination sample set according to a preset performance test type, and determine a target fine-tuned rich element combination according to a test evaluation result. BRIEF DESCRIPTION OF DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative effort.

[0013] Figure 1 A rich element replacement adaptive test optimization method flowchart provided by the embodiment of the present application;

[0014] Figure 2An element replacement adaptive test optimization system structure schematic diagram provided by an embodiment of the present application.

[0015] Reference signs: rich element combination set obtaining module 10, reverse fine-tuning module 20, sputtering experiment module 30, test evaluation module 40. DETAILED DESCRIPTION

[0016] The present application provides an element replacement adaptive test optimization method and system, which is used to solve the technical problems of low test efficiency, high research and development loss, difficult performance guarantee, and insufficient differentiation of rich element replacement of scarce elements in the prior art.

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0018] Embodiment one, as shown in the present application provides an element replacement adaptive test optimization method, which comprises: Figure 1 Step S100: Extract the performance data of scarce elements and the types of scarce elements in the target element replacement scene information, and perform one-order multi-dimensional screening in the rich element resource library to obtain a to-be-verified rich element combination set.

[0019] Specifically, the element replacement scene is defined, the application field is combined, the mature element combination is defined, that is, the element formula that has been applied in batches in the target scene and has complete performance data; the new element combination to be verified, that is, the new formula planned to be replaced, is also defined, and the core indicators that need to be met by the new combination are sorted out. Then, the mature element combinations similar to the application scene of the new element combination and the core performance requirements are screened out from the production database, the complete production parameters and performance data of these combinations are extracted, and a similar mature combination dataset is constructed. Then, the performance data of the scarce elements and the types of the scarce elements are extracted from the target element replacement scene information, the similar mature combination dataset is taken as a reference, one-order multi-dimensional screening is performed in the rich element resource library from the material dimension and the process dimension, the material similar rich element set and the process similar rich element set are determined and included in the candidate rich element pool, then the elements in the pool are combined in a random number and proportion, and the to-be-verified rich element combination set is obtained after screening.

[0020]

[0021] ​Step S200: performing virtual representation reverse fine-tuning on the to-be-verified rich-element combination set based on double targets to obtain a to-be-verified fine-tuned rich-element combination set, wherein the double targets include that the performance similarity between the to-be-verified rich-element combination and the scarce element meets a preset performance requirement and the distribution concentration between virtual representations of the to-be-verified rich-element combination set is less than a preset distribution concentration requirement.

[0022] Specifically, the virtual representation reverse fine-tuning is performed on the to-be-verified rich-element combination set based on the double targets to obtain the to-be-verified fine-tuned rich-element combination set. The double targets are that the performance similarity between the to-be-verified rich-element combination and the scarce element meets the preset performance requirement to ensure that the performance after replacement does not deviate from the core standard, and that the distribution concentration between virtual representations of the to-be-verified rich-element combination set is lower than the preset distribution concentration requirement to avoid too similar combinations and retain the possibility of differentiation. In specific operation, it is first determined whether the virtual representation meets the double targets. If not, the target deviation degree is determined, and fine-tuning is performed according to the pre-set adjustment range. After multiple verifications and adjustments, the to-be-verified fine-tuned rich-element combination set is finally obtained.

[0023] Step S300: performing sample preparation and sputtering experiments on the to-be-verified fine-tuned rich-element combination set to obtain a to-be-verified fine-tuned rich-element combination sample set and a sputtering experiment result set.

[0024] Specifically, for each combination in the to-be-verified fine-tuned rich-element combination set, small-batch sample preparation is first performed through vacuum melting and target processing to obtain a corresponding rich-element or alloy test target. Then, under standardized process conditions, sputtering experiments are performed on each prepared sample to test the deposition rate, density and composition uniformity of the film layer, and various data in the experiment process are recorded. After the entire to-be-verified fine-tuned rich-element combination set is traversed, all prepared samples are integrated to form the to-be-verified fine-tuned rich-element combination sample set, and the sputtering experiment data of each sample are summarized to obtain the sputtering experiment result set.

[0025] Step S400: when the sputtering experiment result set passes the experiment, performing test evaluation on the to-be-verified fine-tuned rich-element combination sample set according to a preset performance test type, and determining a target fine-tuned rich-element combination according to a test evaluation result.

[0026] Specifically, first, the sputtering experiment result set is judged. If the result is that the experiment is passed, that is, the sample meets the basic adaptability of the sputtering process, then according to the preset performance test type including resistivity performance test, transmittance performance test, reflectivity performance test, hardness performance test and adhesion performance test, all the samples of the verified fine-tuning rich element combination to be tested that pass the experiment are comprehensively tested and evaluated. After all the performance tests are completed, the test data is comprehensively weighted and scored, and the sample with the highest comprehensive score that meets all performance indicators is selected. Finally, the rich element combination corresponding to the sample is determined as the target fine-tuning rich element combination.

[0027] In a possible implementation manner, step S400 further includes:

[0028] Step S410: The preset performance test type includes resistivity performance test, transmittance performance test, reflectivity performance test, hardness performance test and adhesion performance test.

[0029] Specifically, the preset performance test type covers multi-dimensional key indicator detection. The resistivity performance test is used to measure the material conductivity, and the resistance characteristics of the current in the material are determined by the instrument. The transmittance performance test focuses on the degree of light transmission of the material, and detects the proportion of specific wavelength light passing through the material. The reflectivity performance test measures the proportion of reflected light on the surface of the material, reflecting the light reflection characteristics of the material. The hardness performance test determines the ability of the material to resist local deformation such as indentation and scratch by using corresponding test means. The adhesion performance test is used to evaluate the firmness of the film layer and the substrate, and to determine whether it is easy to fall off.

[0030] In a possible implementation manner, step S100 further includes:

[0031] Step S110: First-order multi-dimensional filtering is performed on the rich element resource library from the material dimension and the process dimension respectively, with the performance data of the scarce element and the type of the scarce element as constraints, to determine a material similar rich element set and a process similar rich element set.

[0032] Step S120: The material similar rich element set and the process similar rich element set are added to the candidate rich element pool.

[0033] Step S130: A candidate rich element combination set is obtained by performing random number and random proportion of candidate rich element combinations on the candidate rich element pool.

[0034] Step S140: The candidate rich element combination set is subjected to step-by-step directional iterative filtering to obtain the verified rich element combination set.

[0035] Specifically, the performance data of the scarce elements and the type of the scarce elements are used as constraints to carry out first-order multi-dimensional screening on the rich element resource library from the material dimension and the process dimension respectively. In the material dimension, a material-level attention mechanism is used to establish the degree of kinship: first, based on the periodic table, the period and group where the scarce elements are located are determined, and the period difference between the rich elements and the scarce elements is counted, such as the same period difference being 0, the adjacent period difference being 1, and the interval period difference increasing according to the actual interval number; the group difference, such as the same main group difference being 0, the adjacent main group difference being 1, and the secondary group and the main group being calculated according to the group serial number difference, and the distance weight is set to 0.3, and the group difference weight is set to 0.4, because the group has a more significant impact on the chemical properties of the elements; then, through the B3LYP functional in the density functional theory (DFT) combined with the 6-31G(d) basis set, the atomic structure of the rich elements and the scarce elements is simulated to obtain the valence electron layer structure, such as the number of valence electrons, the electron sub-layer distribution, the orbital hybridization type, such as sp³ and sp² hybridization, and other electronic structure parameters, the degree of coincidence of the valence electron arrangement of the two is obtained through calculation, such as the same valence electron number ratio and the sub-layer electron distribution matching ratio to obtain the electron arrangement matching degree, and the electronic cloud distribution similarity is obtained by analyzing the overlap area ratio of the electronic cloud contour map; then, a material-level attention weight is introduced, and the electronic structure similarity is assigned a weight of 0.5, and the adjacentness of the periodic table is quantified with a weight of 0.5, the weight is trained and optimized through historical element replacement data to ensure that the correlation coefficient of the actual material property matching degree is greater than or equal to 0.85, the electron arrangement matching degree, the electronic cloud distribution similarity and the quantification results of the period difference and the group difference are weighted and fused, wherein the period difference quantization value = 1 / (1+period difference), and the group difference quantization value = 1 / (1+group difference), the formula is kinship = 0.5×(0.6×electron arrangement matching degree+0.4×electron cloud distribution similarity)+0.5×(0.3×period difference quantization value+0.4×group difference quantization value), and the rich elements with a kinship greater than or equal to 0.7 are screened out, the threshold is verified through multiple material replacement experiments, which can ensure that the selected rich elements have similar chemical activity, bonding ability and other core material properties to the scarce elements, forming a material similar rich element set; in the process dimension, the process-level attention mechanism is used to calculate the processing feasibility: first, the thermodynamic and process parameters of the rich elements and the scarce elements are measured, the melting point and the boiling point are measured by differential scanning calorimetry, the vapor pressure is measured by static method, and the sputtering yield is calculated by the ratio of the target mass loss per unit time to the sputtering power in the sputtering experiment, the difference value of each parameter is normalized, the normalized value = 1-|rich element parameter value-scarse element parameter value| / scarse element parameter value, and the similarity threshold is set, the normalized value of the melting point and the boiling point is greater than or equal to 0.8, the normalized value of the vapor pressure is greater than or equal to 0.75, and the normalized value of the sputtering yield is greater than or equal to 0.7; then, the similarity of the smelting process, such as melting fluidity, cooling crystallization rate, sputtering, such as target etching uniformity, thin film deposition rate, and other process behaviors is quantified and scored, the similarity score of each process behavior is calculated by weighting the parameter normalized value, and the smelting behavior weight is set to 0.4, the sputtering behavior weight is set to 0.6, because sputtering is the core process of coating target material manufacturing, finally the process level attention weight is used to distribute 0.7 weight to the core process behavior similarity score and 0.3 weight to the non-core process behavior score, the weight is calibrated through process stability test data, the processing feasibility is obtained by weighting and fusing each score, the element-rich elements with processing feasibility greater than or equal to 0.72 are screened out to determine the process similar element-rich element set.

[0036] After the determination of the material similar element-rich element set and the process similar element-rich element set, all the element-rich elements in the two sets are uniformly added to the candidate element-rich element pool according to the relevant information such as the element identifier, the material dimension affinity value, and the process dimension processing feasibility value, so that the candidate element-rich element pool can integrate the element-rich elements screened out from the material and the process, and provide basic resource reserves for subsequent element-rich element combination operations.

[0037] For the candidate element-rich element pool, a random algorithm is set to randomly combine the element-rich elements in the pool, such as combinations of 2 to 5 kinds with random proportions. For example, 3 kinds of element-rich elements may be randomly selected for combination in a ratio of 1:2:1, or 2 kinds of element-rich elements may be randomly selected for combination in a ratio of 3:1, to generate various element-rich element combinations, and finally obtain a candidate element-rich element combination set containing numerous different element-rich element combinations, thereby enriching the sample quantity and type for subsequent screening.

[0038] After obtaining the candidate element-rich element combination set, first, the candidate element-rich element combination set is screened based on the element-rich element combination screening space, and the element-rich element combination screening space is constructed with the process similarity as the horizontal coordinate axis and the material similarity as the vertical coordinate axis. Each candidate element-rich element combination corresponds to a coordinate point in the space. Then, a preset process similarity threshold and a preset material similarity threshold are used as constraints to delimit an element-rich element combination screening subspace, and only the candidate element-rich element combinations whose coordinate points fall within the subspace are retained. Subsequently, stepwise directional iterative screening is performed within the screening subspace: first, the center coordinate point and the optimal coordinate point of the screening subspace, i.e., the coordinate point with the highest process similarity and material similarity, are extracted. The direction from the center coordinate point to the optimal coordinate point is taken as the iteration direction, the mean shift iteration is performed on the center coordinate point according to a preset iteration step, and the iteration coordinate point is determined. The iteration neighborhood density of the center coordinate point is continuously judged to be greater than or equal to the iteration neighborhood density of the iteration coordinate point. If not, the iteration continues until the preset iteration number is reached or the iteration neighborhood density of this iteration is less than the iteration neighborhood density of the last iteration. Finally, based on the target coordinate point obtained through iteration and the preset iteration step, the required combinations are extracted from the candidate element-rich element combination set to form a to-be-verified element-rich element combination set.

[0039] In one possible implementation manner, step S140 further includes:

[0040] Step S141: constructing a rich-element combination screening space based on the candidate rich-element combination set, wherein the horizontal coordinate axis of the rich-element combination screening space is process similarity, the vertical coordinate axis is material similarity, and each coordinate point corresponds to a candidate rich-element combination.

[0041] Step S142: constructing a rich-element combination screening subspace in the rich-element combination screening space, with a preset process similarity threshold and a preset material similarity threshold as constraints.

[0042] Step S143: performing step-by-step directional iterative screening in the rich-element combination screening subspace to obtain the set of to-be-verified rich-element combinations.

[0043] Specifically, based on the previously determined candidate rich-element combination set, a rich-element combination screening space for screening analysis is constructed. The screening space is a two-dimensional space, with the horizontal coordinate axis representing process similarity, and the value of process similarity being derived from the quantified processing feasibility of each candidate rich-element combination and the scarce element in the process dimension. The processing feasibility is obtained by weighting and fusing the melting behavior similarity and sputtering behavior similarity of the rich element and the scarce element. The melting behavior similarity is obtained by measuring the thermodynamic parameters such as melting point, boiling point, and heat of fusion, calculating the parameter difference normalization value, and setting the score according to the threshold. The sputtering behavior similarity is obtained by measuring parameters such as vapor pressure and sputtering yield, calculating the parameter difference normalization value, and setting the score according to the threshold. The vertical coordinate axis represents material similarity, and the value of material similarity corresponds to the quantified degree of kinship of each candidate rich-element combination and the scarce element in the material dimension. The degree of kinship is obtained by weighting and fusing the periodic table proximity and electronic structure similarity of the rich element and the scarce element. The periodic table proximity is obtained by counting the period difference and group difference and setting the distance weight to calculate the base value. The electronic structure similarity is obtained by comparing the electron arrangement, calculating the electron arrangement matching degree, and analyzing the electronic cloud distribution similarity by quantum chemical method. Each candidate rich-element combination will correspond to a unique coordinate point in this two-dimensional screening space according to its corresponding process similarity value and material similarity value, so that all candidate rich-element combinations are presented in the form of coordinate points in the space, providing intuitive and quantifiable analysis basis for subsequent screening operations.

[0044] First, the preset process similarity threshold and the preset material similarity threshold are determined, wherein the preset process similarity threshold corresponds to the minimum qualified standard of the process similarity of the candidate rich-element combination and the feasibility of processing the scarce element, and the preset material similarity threshold corresponds to the minimum qualified standard of the material similarity of the candidate rich-element combination and the affinity of the scarce element. Then, in the rich-element combination screening space, all coordinate points whose process similarity values on the horizontal coordinate axis are greater than or equal to the preset process similarity threshold and whose material similarity values on the vertical coordinate axis are greater than or equal to the preset material similarity threshold are screened out, and the region in which the coordinate points meet the two threshold requirements is defined as a rich-element combination screening subspace, so that only the candidate rich-element combinations that meet the basic qualified standards of process similarity and material similarity are retained in the rich-element combination screening subspace, thereby narrowing the range and focusing on the target in the subsequent further iteration.

[0045] From the constructed rich-element combination screening subspace, the center coordinate point and the optimal coordinate point of the subspace are extracted, wherein the center coordinate point is a point determined by the average process similarity and the average material similarity of all coordinate points corresponding to the candidate rich-element combinations in the subspace, and the optimal coordinate point is a coordinate point in which the process similarity and the material similarity reach the highest values. Then, the direction from the center coordinate point to the optimal coordinate point is taken as the iteration direction, and the mean shift iteration operation is performed on the center coordinate point according to the preset iteration step, so as to determine the coordinate point after iteration. Then, it is judged whether the iteration neighborhood density of the initial center coordinate point is greater than or equal to the iteration neighborhood density of the coordinate point after iteration. The iteration neighborhood density refers to the number of coordinate points corresponding to other candidate rich-element combinations within a certain range around the coordinate point. If the iteration neighborhood density of the initial center coordinate point is not greater than or equal to the iteration neighborhood density of the coordinate point after iteration, the mean shift iteration is performed again on the coordinate point after iteration along the same iteration direction, and the process is repeated until the preset iteration stop condition is met. The preset iteration stop condition includes reaching the preset number of iterations, or the iteration neighborhood density of the coordinate point obtained in a certain iteration is less than the iteration neighborhood density of the coordinate point obtained in the last iteration. When the iteration stop condition is met, the final target coordinate point is determined, and then all coordinate points within the iteration step range from the target coordinate point are extracted from the rich-element combination screening subspace based on the target coordinate point and the preset iteration step. The candidate rich-element combinations corresponding to these coordinate points jointly constitute the to-be-verified rich-element combination set.

[0046] In one possible implementation manner, the step S143 further includes:

[0047] The step S1431 includes extracting the center coordinate point and the optimal coordinate point of the rich-element combination screening subspace.

[0048] Step S1432: Mean shift iteration is performed on the center coordinate point according to a preset iteration step size and in the direction from the center coordinate point to the optimal coordinate point, to determine an iteration coordinate point.

[0049] Step S1433: It is determined whether the iteration neighborhood density of the center coordinate point is greater than or equal to the iteration neighborhood density of the iteration coordinate point. If not, the iteration coordinate point is subjected to mean shift iteration in the iteration direction until a preset iteration stop condition is met, and a target coordinate point is obtained.

[0050] Step S1434: Based on the target coordinate point and the preset iteration step size, the rich element combination screening subspace is extracted, and the to-be-verified rich element combination set is obtained.

[0051] Specifically, for the constructed rich element combination screening subspace, the process similarity mean and the material similarity mean of the coordinate points corresponding to all candidate rich element combinations in the subspace are calculated, and the coordinate point corresponding to the two means is determined as the center coordinate point of the screening subspace, so as to reflect the overall distribution center of the candidate combinations in the subspace. Then, according to a preset weight ratio, such as a process similarity weight of 0.6 and a material similarity weight of 0.4, the process similarity and the material similarity of each coordinate point are weighted and calculated to obtain a weighted comprehensive value of each coordinate point, and the coordinate point with the maximum weighted comprehensive value is selected from all coordinate points and determined as the optimal coordinate point of the screening subspace. The point represents the position of the candidate rich element combination with the optimal comprehensive performance of process and material similarity in the subspace.

[0052] The setting rule of the iteration direction is determined, the center coordinate point extracted in the rich element combination screening subspace is taken as the starting point, and the optimal coordinate point is taken as the ending point. The direction between the two points is the direction of this iteration, which can ensure that the iteration process always advances to the area with better comprehensive performance of process and material similarity. Then, according to the preset iteration step size, such as a step size of 0.1 in the process similarity and material similarity dimensions, the mean shift iteration operation is performed on the center coordinate point: the difference between the center coordinate point and the optimal coordinate point in the process similarity and material similarity dimensions is calculated, and then the moving amplitude in the two dimensions of each iteration is determined according to the iteration step size. The center coordinate point is moved along the iteration direction by the amplitude to obtain a new coordinate point. Finally, the effectiveness of the new coordinate point is confirmed by verifying whether the process similarity and the material similarity of the new coordinate point meet the adjustment logic corresponding to the iteration step size, and the effective new coordinate point is finally determined as the iteration coordinate point.

[0053] The definition of the iterative neighborhood density is determined, that is, a neighborhood range is determined around the coordinate point, such as a region with a process similarity and a material similarity of 0.05 as the radius and the coordinate point as the center, the number of the coordinate points corresponding to other candidate element-rich combinations contained in the neighborhood is counted, and the number is the iterative neighborhood density of the coordinate point. Then, the iterative neighborhood densities of the initial center coordinate point and the iterative coordinate point are compared: if the iterative neighborhood density of the center coordinate point is greater than or equal to the iterative neighborhood density of the iterative coordinate point, it is indicated that the concentration degree of the candidate combinations in the neighborhood is not improved in the current iterative direction, and the iteration in the current direction can be paused; if the iterative neighborhood density of the center coordinate point is less than the iterative neighborhood density of the iterative coordinate point, it is indicated that a more optimal candidate combination concentration region can be obtained by continuing to advance in the iterative direction, and the current iterative coordinate point is taken as a new starting point, the mean shift iteration is performed again according to the previously determined iterative direction and the preset iterative step length, and a new iterative coordinate point is generated. Then, the neighborhood density comparison and the iteration operation are repeated until the preset iteration stop condition is met, the maximum number of iterations is reached, or the iterative neighborhood density of the new coordinate point generated in the iteration is less than the iterative neighborhood density of the last iterative coordinate point, at which time the iteration is stopped, and the coordinate point obtained in the last iteration is determined as the target coordinate point.

[0054] On the horizontal coordinate axis of the element-rich combination screening subspace, the extraction interval of the process similarity is formed by extending the distance corresponding to the preset iterative step length in the two directions of increasing and decreasing the value of the process similarity corresponding to the target coordinate point as the center; on the vertical coordinate axis, the extraction interval of the material similarity is also formed by extending the distance corresponding to the preset iterative step length in the two directions of increasing and decreasing the value of the material similarity corresponding to the target coordinate point as the center. The extraction intervals of the two coordinate axes jointly form a rectangular extraction region, which is the target range of the subspace extraction this time. Then, all the coordinate points corresponding to the candidate element-rich combinations in the element-rich combination screening subspace are traversed, and it is judged whether the process similarity of each coordinate point falls within the process similarity extraction interval and whether the material similarity falls within the material similarity extraction interval. Finally, the candidate element-rich combinations corresponding to all the coordinate points that meet the requirements of the two intervals are screened out, these combinations are sorted and summarized, and finally the set of element-rich combinations to be verified is formed.

[0055] In one possible implementation manner, step S1433 further includes:

[0056] The preset iteration stop condition is a preset number of iterations and / or the iterative neighborhood density of this iteration is less than the iterative neighborhood density of the last iteration.

[0057] Specifically, the preset iteration stopping condition includes two triggerable cases, and the iteration process is terminated when any of the two cases is met. The first case is a preset iteration number, i.e., when the number of iteration operations on the mean shift of the coordinate points in the rich element combination screening subspace reaches a preset fixed number, the iteration is stopped regardless of the current iteration effect, so as to ensure that the iteration process will not be performed indefinitely. The second case is the change of the iteration neighborhood density, i.e., the iteration neighborhood density of the coordinate points obtained in this iteration is compared with the iteration neighborhood density of the coordinate points obtained in the last iteration. If the iteration neighborhood density of this iteration is less than the iteration neighborhood density of the last iteration, it indicates that the concentration of the candidate rich element combinations in the neighborhood cannot be improved by continuing to iterate in the current direction, but instead, it is decreased. At this time, the iteration also needs to be stopped. Through the setting of the two conditions, the resource consumption caused by over-iteration can be avoided, and the effectiveness of the iteration result can be ensured.

[0058] In a possible implementation manner, step S200 further includes:

[0059] Step S210: determining whether the virtual representation of the set of rich element combinations to be verified meets the double target, and if not, determining a target deviation degree.

[0060] Step S220: fine-tuning the set of rich element combinations to be verified according to a preset adjustment mode based on the target deviation degree, to obtain a set of once-adjusted rich element combinations to be verified.

[0061] Step S230: performing double-target verification on the virtual representation of the set of once-adjusted rich element combinations to be verified again, and obtaining a set of fine-tuned rich element combinations to be verified through multiple verification and adjustment.

[0062] Specifically, the performance simulation calculation of the set of rich element combinations to be verified is performed by using a virtual simulation software, similarity data of each combination and the scarce element in the key performance dimensions such as resistivity, transmittance and reflectivity are obtained, and the dispersion coefficients of all combinations in the element ratio and performance parameters are calculated by using a data statistical tool, so as to quantitatively determine the performance similarity and the combination difference in the virtual representation. Then, a preset double-target threshold database is called, the qualified threshold of the performance similarity (such as ≥90%) and the qualified threshold of the combination difference (such as the dispersion coefficient ≥0.1) are extracted, the calculated average value of the performance similarity is compared with the performance threshold, and the combination difference is compared with the difference threshold. If the average value of the performance similarity does not reach the threshold, or the combination difference does not reach the threshold, it is determined that the double target is not met, and at this time, the deviation degree is calculated: the difference between the performance similarity threshold and the actual average value is multiplied by the performance weight (such as 0.6), and the difference between the combination difference threshold and the actual value is multiplied by the difference weight (such as 0.4), and the weighted result is the target deviation degree.

[0063] The determined target deviation degree data is called to determine whether the deviation mainly comes from the failure of performance similarity or the failure of combination difference, and a preset adjustment mode is extracted from a preset parameter library. The element proportion of the to-be-verified rich element combination set is adjusted according to a fixed preset adjustment amplitude. For example, the key elements affecting the performance are adjusted by ±5% amplitude, and the auxiliary elements affecting the difference are adjusted by ±3% amplitude. Subsequently, for each combination in the to-be-verified rich element combination set, if the target deviation mainly comes from the insufficient performance similarity, the elements with high performance correlation to the scarce elements, such as metal elements affecting the resistivity, are adjusted according to the adjustment amplitude to increase or decrease the proportion, so as to improve the performance similarity. If the deviation mainly comes from the too low combination difference, the elements with similar proportions in each combination are adjusted according to the adjustment amplitude difference to expand the element proportion gap between combinations and enhance the difference. After the element proportion of all combinations is finely adjusted, the adjusted combinations are integrated to form a to-be-verified once-adjusted rich element combination set.

[0064] The double-target verification logic and implementation means are used. The performance similarity of each combination in the to-be-verified once-adjusted rich element combination set and the scarce elements is calculated by a virtual simulation software. The difference (dispersion coefficient) between combinations is analyzed by a data statistical tool, and is compared with a preset performance similarity threshold and a difference threshold, respectively. If the verification result shows that the set has met the double targets, that is, the performance similarity meets the standard and the difference meets the requirement, the set is directly determined as a to-be-verified fine-adjusted rich element combination set. If the double targets are not met, the deviation degree calculation process is restarted. After a new target deviation degree is determined, the element proportion is adjusted according to the preset amplitude. The to-be-verified once-adjusted rich element combination set is finely adjusted again to obtain a to-be-verified twice-adjusted rich element combination set. Then the above-mentioned “virtual representation calculation→double-target verification→deviation degree calculation→element proportion fine adjustment” cycle operation is repeated until the rich element combination set after a certain adjustment passes the double-target verification. At this time, the cycle is stopped, and the set after the adjustment is finally determined as the to-be-verified fine-adjusted rich element combination set.

[0065] In a possible implementation manner, step S220 further includes:

[0066] The preset adjustment mode is to adjust the element proportion of the to-be-verified rich element combination set according to the preset adjustment amplitude.

[0067] Specifically, in the adaptive test optimization process of element replacement, when fine-tuning of the to-be-verified rich-element combination set is needed, the preset adjustment mode adopted is to adjust the element proportion according to a preset adjustment amplitude. First, the preset adjustment amplitude is extracted from the system parameter configuration, such as setting the adjustment amplitude of ±5% for elements affecting the core performance and setting the adjustment amplitude of ±3% for auxiliary elements, to clearly limit the adjustment range of different types of elements. Then, in combination with the target deviation degree of the to-be-verified rich-element combination set, the deviation source is judged. If the deviation is caused by the performance of a certain type of element, the proportion of the element is increased or decreased by the preset amplitude, for example, when the performance similarity is insufficient, the proportion of the element with high performance correlation with the scarce element is increased; if the deviation is caused by too low combination difference, the elements with similar proportions in each combination are differentiated by the preset amplitude, such as increasing the proportion of a certain element in some combinations and decreasing the proportion of the element in some combinations. In the adjustment process, the sum of the element proportions of each combination is always ensured to be 100%, avoiding the problem of proportion imbalance, and finally the element proportion fine-tuning of all to-be-verified rich-element combinations is completed, forming an adjusted rich-element combination set.

[0068] In one possible implementation manner, step S400 further includes:

[0069] Step S410: Obtain a production feedback monitoring window.

[0070] Step S420: Material preparation is performed on the target fine-tuning rich-element combination in the production feedback monitoring window, and quality evaluation is performed on the prepared material. If the quality evaluation result is less than the preset quality requirement, a test warning instruction is generated.

[0071] Specifically, the definition basis of the production feedback monitoring window is determined, and the time range and production batch range of the monitoring are determined in combination with the target fine-tuning rich-element combination corresponding to the target material production characteristics. The time range is usually selected to cover a period of time covering the complete production cycle, such as 3 consecutive working days, to ensure that each key stage of material preparation is included; the production batch range is set to a number with statistical significance, such as 6-10 consecutive production batches, to avoid the contingency of single batch data. Subsequently, by analyzing the key nodes in the production process, such as alloy batching, casting, rolling, and finished product detection, the production links and data types that need to be tracked in the monitoring window are determined, such as the process parameters of each link, the quality detection data of semi-finished products and finished products, and finally the monitoring window that can accurately capture the production feedback information of the target rich-element combination is determined, providing a clear data collection boundary for subsequent quality evaluation.

[0072] During the determined production feedback monitoring window, the material preparation work is carried out using the target fine-tuning rich element combination according to the standard process of target material manufacturing, from alloy batching, casting, rolling to forming, each step strictly follows the process parameters matched with the rich element combination. After the material preparation is completed, the quality of each batch of finished products is comprehensively detected according to the preset quality evaluation indicators such as the density, purity, surface flatness and internal density of the target material, and the detection data is recorded and the quality evaluation result is calculated. The calculated quality evaluation result is compared with the pre-set quality requirement, if the quality evaluation result of one or more batches of materials does not meet the pre-set quality requirement, it is determined that the test optimization of the target fine-tuning rich element combination before exists deviation, at this time, the test warning instruction needs to be generated, and it is clearly prompted to return to the previous step of element replacement adaptability test, and the test optimization work such as virtual characterization reverse fine-tuning and sputtering experiment is carried out again, so as to ensure that the quality of the material prepared subsequently meets the production requirements.

[0073] In the embodiment two, based on the same inventive concept as the adaptability test optimization method of replacing one element in the foregoing embodiment, as shown in the embodiment two, the present application provides an adaptability test optimization system of replacing elements, and the system and method embodiments in the present application are based on the same inventive concept. Wherein, the system comprises: Figure 2

[0074] The rich element combination set acquisition module 10 is used to extract the performance data of the scarce element and the type of the scarce element in the target element replacement scene information, and perform one-dimensional multi-dimensional screening in the rich element resource library to obtain a set of verified rich element combinations.

[0075] The reverse fine-tuning module 20 is used to perform virtual characterization reverse fine-tuning on the set of verified rich element combinations based on double targets, to obtain a set of verified fine-tuned rich element combinations, wherein the double targets include that the similarity of the verified rich element combination and the performance of the scarce element meets the pre-set performance requirement and the distribution centralization between the virtual characterization of the set of verified rich element combinations is less than the pre-set distribution centralization requirement.

[0076] The sputtering experiment module 30 is used to traverse the set of verified fine-tuned rich element combinations to perform sample preparation and sputtering experiment, to obtain a set of verified fine-tuned rich element combination samples and a set of sputtering experiment results.

[0077] The test evaluation module 40 is used to perform test evaluation on the set of verified fine-tuned rich element combination samples according to the pre-set performance test type when the set of sputtering experiment results is passed, and determine the target fine-tuned rich element combination according to the test evaluation result.

[0078] Further, the system is also used to implement the following functions:

[0079] ​The preset performance test types include resistivity performance test, transmittance performance test, reflectance performance test, hardness performance test and adhesion performance test.

[0080] Further, the system is further used to implement the following functions:

[0081] With the rare element performance data and the rare element type as constraints, first-order multi-dimensional screening is performed on the rich element resource library from the material dimension and the process dimension respectively to determine a material similar rich element set and a process similar rich element set; the material similar rich element set and the process similar rich element set are added into a candidate rich element pool; random quantity and random proportion of candidate rich element combination are performed on the candidate rich element pool to obtain a candidate rich element combination set; stepwise directional iterative screening is performed on the candidate rich element combination set to obtain the to-be-verified rich element combination set.

[0082] Further, the system is further used to implement the following functions:

[0083] Based on the candidate rich element combination set, a rich element combination screening space is constructed, wherein a horizontal coordinate axis of the rich element combination screening space is process similarity, a vertical coordinate axis is material similarity, and each coordinate point corresponds to a candidate rich element combination; in the rich element combination screening space, with a preset process similarity threshold and a preset material similarity threshold as constraints, a rich element combination screening subspace is constructed; stepwise directional iterative screening is performed in the rich element combination screening subspace to obtain the to-be-verified rich element combination set.

[0084] Further, the system is further used to implement the following functions:

[0085] A center coordinate point and an optimal coordinate point of the rich element combination screening subspace are extracted; with the center coordinate point to the optimal coordinate point as an iterative direction, mean shift iteration is performed on the center coordinate point according to a preset iterative step length to determine an iterative coordinate point; it is judged whether the iterative neighborhood density of the center coordinate point is greater than or equal to the iterative neighborhood density of the iterative coordinate point, if not, the mean shift iteration will be continued on the iterative coordinate point according to the iterative direction until a preset iteration stop condition is met to obtain a target coordinate point; based on the target coordinate point and the preset iterative step length, the rich element combination screening subspace is extracted to obtain the to-be-verified rich element combination set.

[0086] Further, the system is further used to implement the following functions:

[0087] The preset iteration stop condition is a preset iteration number and / or the iterative neighborhood density of this iteration is less than the iterative neighborhood density of the last iteration.

[0088] Further, the system is further used to realize the following functions:

[0089] If the virtual representation of the to-be-verified rich-element combination set does not satisfy the double target, a target deviation degree is determined, the to-be-verified rich-element combination set is fine-tuned according to a preset adjustment mode based on the target deviation degree, a to-be-verified once-adjusted rich-element combination set is obtained, the virtual representation of the to-be-verified once-adjusted rich-element combination set is verified again according to the double target, and the to-be-verified fine-tuned rich-element combination set is obtained through multiple verification and adjustment.

[0090] Further, the system is further used to realize the following functions:

[0091] The preset adjustment mode is to adjust the element proportion of the to-be-verified rich-element combination set according to a preset adjustment amplitude.

[0092] Further, the system is further used to realize the following functions: a production feedback monitoring window is obtained, the target fine-tuned rich-element combination is prepared in the production feedback monitoring window, the prepared material is evaluated in quality, and if the quality evaluation result is less than a preset quality requirement, a test early warning instruction is generated.

[0093] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. And the above describes a specific embodiment of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0094] The above only describes the preferred embodiments of the present application, and does not limit the present application. 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.

[0095] The specification and drawings are only exemplary of the present application, and should be considered to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. A method of adaptive test optimization for element replacement, characterized in that, The method comprises: extracting the performance data and type of the scarce elements in the target element replacement scenario information, and performing one-dimensional multi-dimensional screening in the rich element resource library to obtain a set of rich element combinations to be verified; based on double targets, virtually representing and reverse-tuning the set of rich element combinations to be verified to obtain a set of rich element combinations to be verified after tuning, wherein the double targets include that the similarity of the set of rich element combinations to be verified to the performance of the scarce elements meets a preset performance requirement and the distribution concentration between the virtual representation of the set of rich element combinations to be verified is less than a preset distribution concentration requirement; iterating through the set of rich element combinations to be verified after tuning to perform sample preparation and sputtering experiments to obtain a set of samples of rich element combinations to be verified after tuning and a set of sputtering experiment results; when the set of sputtering experiment results passes the experiment, testing and evaluating the set of samples of rich element combinations to be verified after tuning according to a preset performance test type, and determining a target rich element combination after tuning according to the test evaluation result; extracting the performance data and type of the scarce elements in the target element replacement scenario information, and performing one-dimensional multi-dimensional screening in the rich element resource library to obtain a set of rich element combinations to be verified, comprising: performing one-dimensional multi-dimensional screening on the rich element resource library from the material dimension and the process dimension respectively with the performance data and type of the scarce elements as constraints to determine a set of material-similar rich elements and a set of process-similar rich elements; adding the set of material-similar rich elements and the set of process-similar rich elements into a candidate rich element pool; obtaining a set of candidate rich element combinations by randomly selecting a number and a proportion of candidate rich element combinations from the candidate rich element pool; obtaining the set of rich element combinations to be verified by stepwise directional iterative screening on the set of candidate rich element combinations.

2. A method for adaptive test optimization by element replacement as claimed in claim 1, characterized in that, The preset performance test type includes resistivity performance test, transmittance performance test, reflectivity performance test, hardness performance test, and adhesion performance test.

3. The method of adaptive test optimization by element substitution of claim 1, wherein, obtaining the set of rich element combinations to be verified by stepwise directional iterative screening on the set of candidate rich element combinations, comprising: constructing a rich element combination screening space based on the set of candidate rich element combinations, wherein the horizontal coordinate axis of the rich element combination screening space is process similarity, the vertical coordinate axis is material similarity, and each coordinate point corresponds to a candidate rich element combination; constructing a rich element combination screening subspace in the rich element combination screening space with a preset process similarity threshold and a preset material similarity threshold as constraints; obtaining the set of rich element combinations to be verified by stepwise directional iterative screening in the rich element combination screening subspace.

4. A method of adaptive test optimization for element replacement as claimed in claim 3, wherein, obtaining the set of rich element combinations to be verified by stepwise directional iterative screening in the rich element combination screening subspace, comprising: extracting the center coordinate point and the optimal coordinate point of the rich element combination screening subspace; taking the center coordinate point to the optimal coordinate point as the iteration direction, and performing mean shift iteration on the center coordinate point according to a preset iteration step to determine an iteration coordinate point; determining whether the iteration neighborhood density of the center coordinate point is greater than or equal to the iteration neighborhood density of the iteration coordinate point, and if not, the mean shift iteration of the iteration coordinate point will continue to be performed in the iteration direction until a preset iteration stop condition is met, and a target coordinate point is obtained; based on the target coordinate point and the preset iteration step, extracting the rich element combination screening subspace to obtain the to-be-verified rich element combination set.

5. A method of adaptive test optimization for element replacement as claimed in claim 4, wherein, The preset iteration stop condition is a preset iteration number and / or the iteration neighborhood density of this iteration is less than the iteration neighborhood density of the last iteration.

6. The method of adaptive test optimization by element substitution of claim 1, wherein, Based on the dual target, the to-be-verified rich element combination set is virtually represented and fine-tuned to obtain a to-be-verified fine-tuned rich element combination set, wherein the dual target includes that the to-be-verified rich element combination and the performance similarity of the scarce element meet a preset performance requirement and the distribution concentration between the virtual representation of the to-be-verified rich element combination set is less than a preset distribution concentration requirement, including: determining whether the virtual representation of the to-be-verified rich element combination set meets the dual target, and if not, determining a target deviation degree; based on the target deviation degree, the to-be-verified rich element combination set is fine-tuned according to a preset adjustment mode to obtain a to-be-verified once-adjusted rich element combination set; the virtual representation of the to-be-verified once-adjusted rich element combination set is verified again, and after multiple verification adjustments, the to-be-verified fine-tuned rich element combination set is obtained.

7. A method of adaptive test optimization for element replacement as defined in claim 6, wherein, The preset adjustment mode is to adjust the element ratio of the to-be-verified rich element combination set according to a preset adjustment amplitude.

8. The method of adaptive test optimization by element substitution of claim 1, wherein, including: obtaining a production feedback monitoring window; material preparation is performed on the target fine-tuned rich element combination in the production feedback monitoring window, and quality evaluation is performed on the prepared material, and if the quality evaluation result is less than a preset quality requirement, a test warning instruction is generated.

9. An adaptive test optimization system for element replacement, characterized by, The system is used to implement the element replacement adaptive test optimization method of any one of claims 1-8, and the system comprises: A rich element combination set acquisition module is configured to extract the performance data and type of the scarce element in the target element replacement scene information, and perform one-dimensional multi-dimensional screening in the rich element resource library to obtain a to-be-verified rich element combination set. A reverse fine-tuning module is configured to virtually represent and fine-tune the to-be-verified rich element combination set based on a dual target to obtain a to-be-verified fine-tuned rich element combination set, wherein the dual target includes that the to-be-verified rich element combination and the performance similarity of the scarce element meet a preset performance requirement and the distribution concentration between the virtual representation of the to-be-verified rich element combination set is less than a preset distribution concentration requirement. A sputtering experiment module is configured to traverse the to-be-verified fine-tuned rich element combination set to perform sample preparation and sputtering experiments, and obtain a to-be-verified fine-tuned rich element combination sample set and a sputtering experiment result set. A test evaluation module is configured to, when the sputtering experiment result set passes the experiment, perform test evaluation on the to-be-verified fine-tuned rich element combination sample set according to a preset performance test type, and determine a target fine-tuned rich element combination according to the test evaluation result.

Images