Quality detection method, device and equipment of nano material and storage medium

Abstract

The invention provides a nano material quality detection method, device and equipment and a storage medium, and relates to the field of material quality detection. The quality detection model is established by combining the first refractive indexes of the nano material under different incident light irradiation and the consistency degree value, one uniformity value and one purity value of the nano material, and the established and generated quality detection model can directly realize quality detection of the nano material based on the refractive indexes, so that the quality detection efficiency is improved. And powerful support is provided for quality control and performance optimization of the nano material.

Description

Technical Field

[0001] This application relates to the field of material quality inspection, and particularly to a method, device, equipment, and storage medium for quality inspection of nanomaterials. Background Art

[0002] The refractive index is an important parameter characterizing the optical properties of nanomaterials, and is closely related to physical properties such as the compositional consistency, structural uniformity, and purity of nanomaterials. Therefore, quality inspection based on the refractive index has become an effective means for nanomaterial detection. However, current detection methods mostly rely on single physical measurements or simple empirical models, and it is difficult to comprehensively capture the complex relationships between the refractive index and uniformity, purity, and structural consistency. Especially when facing diverse nanomaterial systems, both the accuracy and universality are insufficient. These methods often ignore the multi-dimensional characteristics of refractive index data and lack effective identification of abnormal data, resulting in limited reliability of the quality assessment results. Summary of the Invention

[0003] In view of this, this application provides a method, device, equipment, and storage medium for quality inspection of nanomaterials, and the main purpose is to provide a method for quality inspection of nanomaterials. Using this model, it is possible to quickly determine the consistency degree value, uniformity value, and purity value representing the quality parameters of nanomaterials based on the refractive index of nanomaterials.

[0004] To achieve the above object, a first aspect of this application discloses a method for quality inspection of nanomaterials, and the method includes:

[0005] Obtain the first refractive index of the nanomaterial when irradiated by a target incident light at different wavelengths. The obtained first refractive index and the wavelength of the target incident light form a first data matrix. The nanomaterial includes at least one first consistency degree value, at least one first uniformity value, and at least one first purity value. In the first data matrix, the first refractive index and the wavelength of the target incident light correspond one by one;

[0006] Use the principal component analysis method to reduce the dimension of the first data matrix. Select a target wavelength range in the reduced first data matrix. The wavelength of the target incident light included in the target wavelength range and the first refractive index corresponding to the wavelength of the target incident light are used to form a second data matrix. The information content value of the second data matrix is the highest. The first uniformity value, the first purity value, and the first refractive index corresponding to the second data matrix are respectively used as the second uniformity value, the second purity value, and the second refractive index;

[0007] Based on the support vector regression algorithm, a quality detection model representing the non-linear relationship between the second data matrix, the second consistency degree value, the second uniformity value, and the second purity value is established. When the target refractive index of the nanomaterial is input into the quality detection model, the quality detection model outputs the target consistency degree value, the target uniformity value, and the target purity value of the nanomaterial.

[0008] In the second aspect of the embodiments of the present application, a quality detection device for nanomaterials is provided. The device includes:

[0009] An acquisition module, configured to acquire the first refractive index of the nanomaterial when irradiated by the target incident light at different wavelengths. The acquired first refractive index and the wavelength of the target incident light are combined into a first data matrix. The nanomaterial includes at least one first consistency degree value, at least one first uniformity value, and at least one first purity value. In the first data matrix, the first refractive index and the wavelength of the target incident light correspond one by one;

[0010] A selection module, configured to reduce the dimension of the first data matrix by using the principal component analysis method, select a target wavelength range in the dimension-reduced first data matrix, and use the wavelength of the target incident light included in the target wavelength range and the first refractive index corresponding to the wavelength of the target incident light to form a second data matrix. The second data matrix has the highest information content value. The first uniformity value, the first purity value, and the first refractive index corresponding to the second data matrix are used as the second uniformity value, the second purity value, and the second refractive index, respectively;

[0011] A detection module, configured to establish a quality detection model representing the non-linear relationship between the second data matrix, the second consistency degree value, the second uniformity value, and the second purity value based on the support vector regression algorithm. When the target refractive index of the nanomaterial is input into the quality detection model, the quality detection model outputs the target consistency degree value, the target uniformity value, and the target purity value of the nanomaterial.

[0012] In the third aspect of the embodiments of the present application, an electronic device is provided, including:

[0013] At least one processor; and a memory communicatively connected to the at least one processor. Wherein, the memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to execute the method according to any one of the first aspect disclosed above.

[0014] In a fourth aspect embodiment of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method described in the first aspect is implemented.

[0015] In summary, according to the technical solution disclosed in the present application, a method for quality inspection of nanomaterials is disclosed, including: obtaining the first refractive index of the nanomaterial when irradiated by target incident light at different wavelengths, and the combination of the obtained first refractive index and the wavelength of the target incident light is a first data matrix. The nanomaterial includes at least one first consistency degree value, at least one first uniformity value, and at least one first purity value, and the first refractive index and the wavelength of the target incident light in the first data matrix correspond one by one; using the principal component analysis method to reduce the dimension of the first data matrix, selecting a target wavelength range in the dimension-reduced first data matrix, and using the wavelength of the target incident light included in the target wavelength range and the first refractive index corresponding to the wavelength of the target incident light to form a second data matrix, and the information quantity value of the second data matrix is the highest. The first uniformity value, first purity value, and first refractive index corresponding to the second data matrix are used as the second uniformity value, second purity value, and second refractive index respectively; based on the support vector regression algorithm, a quality inspection model representing the non-linear relationship between the second data matrix, the second consistency degree value, the second uniformity value, and the second purity value is established. When the target refractive index of the nanomaterial is input into the quality inspection model, the quality inspection model outputs the target consistency degree value, target uniformity value, and target purity value of the nanomaterial. The present application combines the first refractive index of the nanomaterial under different incident light irradiations, as well as the consistency degree value, a uniformity value, and a purity value of the nanomaterial to construct a quality inspection model. The constructed quality inspection model can directly perform quality inspection of the nanomaterial based on the refractive index, providing strong support for the quality control and performance optimization of the nanomaterial.

[0016] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features, and advantages of the present application more obvious and understandable, the following specific embodiments of the present application are specifically given. Brief Description of the Drawings

[0017] The drawings here are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application.

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0019] Figure 1 The flowchart of the quality detection method for the nanomaterial provided by the embodiment of the present application is shown;

[0020] Figure 2 The structural diagram of a quality detection device for a nanomaterial provided by the embodiment of the present application is shown. Specific embodiments

[0021] In order to more clearly understand the above-mentioned objects, features and advantages of the present application, the solution of the present application will be further described below. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments may be combined with each other.

[0022] In order to solve the problem that the existing detection methods are difficult to comprehensively capture the complex correlation between the refractive index and the uniformity, purity and structural consistency. The present application provides the following embodiments to solve the above problems:

[0023] This embodiment provides a quality detection method for a nanomaterial, as Figure 1 shown, which is the flowchart of the method of this embodiment. The method of this embodiment may specifically include the following steps:

[0024] Step 101, obtain the first refractive index of the nanomaterial when irradiated by the target incident light at different wavelengths. The obtained first refractive index and the wavelength of the target incident light form a first data matrix. The nanomaterial includes at least one first consistency degree value, at least one first uniformity value and at least one first purity value. In the first data matrix, the first refractive index of the target incident light and the wavelength of the target incident light correspond one by one.

[0025] The consistency degree value mainly refers to the consistency degree value of the nanoparticles in terms of size, shape, structure and chemical properties, etc. The first consistency degree value, the second consistency degree value and the third consistency degree value are different consistency degree values. The uniformity value reflects the distribution uniformity of the grain size in the microstructure of the nanomaterial. Specifically, the uniformity refers to the degree to which the distribution of the grain size in the nanomaterial approaches the ideal state, that is, the standard deviation of the grain size distribution approaches the Hillert limit (0.354). The first uniformity value, the second uniformity value and the third uniformity value are different uniformity values. The purity value represents the content ratio of the target component. The first purity value, the second purity value and the third purity value are different purity values. The consistency degree value, the uniformity and the purity value are used as quality parameters to represent the material quality of the nanomaterial.

[0026] Collect the refractive index data of the nanomaterial under full-band incident light irradiation by an optical measurement device as the first refractive index. Record the first refractive index value for the incident light at each wavelength point, and combine them to generate a first data matrix. The incident lights with different wavelengths irradiated on the nanomaterial are used as the target incident lights. Additionally, there can be multiple nanomaterials irradiated by the incident light, and the multiple nanomaterials can have different consistency degree values, uniformity values, and purity values. In this case, the consistency degree value, uniformity value, and purity value of the nanomaterial are used as the first consistency degree value, the first uniformity value, and the first purity value.

[0027] Extract the target incident lights at multiple wavelength points and the corresponding first refractive indices from the first data matrix, and use a data cleaning method to remove the noise data to obtain a denoised refractive index data matrix. For the denoised refractive index data matrix, calculate the change rate of the first refractive index value corresponding to each wavelength point, and determine the first refractive index fluctuation characteristics corresponding to the incident lights between adjacent wavelength points. If the first refractive index fluctuation characteristics between adjacent wavelength points exceed the preset threshold, supplement the incident lights of the missing wavelengths and the corresponding first refractive indices through an interpolation algorithm to generate a smoothed refractive index data matrix. According to the smoothed refractive index data matrix, use a clustering algorithm to divide the wavelength range to obtain the refractive index distribution pattern of the wavelength range. Analyze the multi-wavelength characteristics of the nanomaterial through the refractive index distribution pattern.

[0028] Specifically, when collecting the first refractive index of the nanomaterial under different target incident light irradiations by an optical measurement device, an ellipsometer can be used to irradiate the nanomaterial at intervals of 1 nm in the wavelength range of 200 nm to 1000 nm and collect the first refractive index to form a first data matrix.

[0029] For example, assume that the refractive index measured at 500 nm is 1.45 and at 501 nm is 1.46. In this way, a two-dimensional matrix is formed, where the horizontal axis is the wavelength of the target incident light and the vertical axis is the first refractive index value, which can comprehensively capture the optical responses of the nanomaterial with the first consistency degree value, the first uniformity value, and the first purity value at different wavelengths and provide a data basis for subsequent analysis.

[0030] When using a data cleaning method to remove the noise data, a moving average filtering method can be used. For example, take the average of 5 adjacent points of the first refractive index data near the incident light at a wavelength of 500 nm to smooth out the mutation values caused by equipment jitter. For example, adjust the abnormal value of 1.50 to a value near 1.45 to obtain a denoised refractive index data matrix. This can significantly improve the data reliability and lay a foundation for accurate analysis. Calculating the refractive index change rate for the denoised refractive index data matrix is an effective means to quantify the fluctuation characteristics.

[0031] For example, in the range of 500 nm to 510 nm, if the refractive index changes from 1.45 to 1.47, the rate of change is (1.47 - 1.45) / 10 = 0.002 / nm.

[0032] Preferably, the rate-of-change threshold can be set to 0.005 / nm. If this value is exceeded, it indicates abnormal fluctuations.

[0033] It should be noted that such fluctuations may be caused by internal defects in the material. Identifying them helps in understanding the material properties. If the fluctuation characteristics exceed the threshold, interpolation algorithms can be used to supplement the data.

[0034] In one embodiment, assuming that the data from 600 nm to 610 nm is missing, the linear interpolation method can be used. Based on 1.48 at 599 nm and 1.50 at 611 nm, it can be deduced that the refractive index at 605 nm is approximately 1.49, generating a smooth refractive-index data matrix. This method can effectively fill the data gaps, improve the continuity of the curve, and facilitate subsequent analysis. Based on the smooth refractive-index data matrix, clustering algorithms can divide the wavelength range.

[0035] For example, using the K-means clustering algorithm, which is an unsupervised machine learning method based on distance. It iteratively divides the data into K clusters to maximize the similarity within the clusters and the differences between the clusters. The K-means algorithm can divide the range from 200 nm to 1000 nm into three intervals: the refractive index is stable at around 1.40 from 200 - 400 nm, gradually rises to 1.50 from 400 - 700 nm, and tends to 1.45 from 700 - 1000 nm. Such a distribution pattern clearly shows the multi-wavelength characteristics of the nanomaterial. This division helps to optimize the application of the material in specific wavelength bands. Analyzing the multi-wavelength characteristics through the refractive-index distribution pattern and generating a data table is the final output of the research.

[0036] Specifically, the average refractive index, change trend, etc. of each wavelength range can be recorded.

[0037] For example, the average refractive index in the range of 400 - 700 nm is 1.47, and the refractive-index change trend is upward. This not only intuitively reflects the optical behavior of the material but also provides data support for the design of optical devices, such as increasing the light transmittance or reflectance in specific wavelength bands. The benefit of this analysis is to convert complex data into practical information and promote the application of nanomaterials in the optical field.

[0038] Step 102: Use the principal component analysis method to reduce the dimension of the first data matrix. Select the target wavelength range from the reduced-dimension first data matrix. The wavelengths of the target incident light included in the target wavelength range and the first refractive indices corresponding to the wavelengths of the target incident light are used to form a second data matrix. The second data matrix has the highest information content value. The first uniformity value, first purity value, and first refractive index corresponding to the second data matrix are respectively used as the second uniformity value, second purity value, and second refractive index.

[0039] Perform a dimension reduction operation on the first data matrix through the principal component analysis method, extract eigenvectors from the reduced-dimension first data matrix to obtain a preliminary feature data set. For the preliminary feature data set, use a standardization processing method to adjust the data distribution to obtain a standardized feature data set. Based on the standardized feature data set, calculate the contribution rate of each eigenvector through variance calculation to obtain a feature contribution rate table. If the contribution rate of a certain eigenvector in the feature contribution rate table is lower than a preset threshold, then eliminate this eigenvector to obtain a refined feature data set. For the refined feature data set, use a clustering algorithm to divide the wavelength range to obtain the target wavelength range. Through the target wavelength range, calculate the mean value of the refractive index data in each range to obtain a wavelength range mean value table. Based on the wavelength range mean value table, extract the wavelength range reflecting the material quality to obtain the final feature data set.

[0040] Specifically, performing a dimension reduction operation on the refractive index data through the principal component analysis method is an important means to extract key information from the first data matrix.

[0041] Exemplarily, assume that the first data matrix contains target incident light from 200nm to 1000nm and the corresponding refractive index values respectively, which can be regarded as a high-dimensional matrix. The principal component analysis will identify the direction in which the eigenvectors change most significantly in the data matrix to form eigenvectors. In a possible implementation, it is found after analysis that the first eigenvector captures 80% of the change, the second captures 15%, and the rest are smaller. In this way, retaining the main eigenvectors can simplify the data. For the preliminary feature data set, it is crucial to use a standardization processing method to adjust the data distribution.

[0042] Specifically, the refractive index values may vary greatly due to different wavelength ranges. For example, it is 1.3 at 200nm and 1.6 at 1000nm. Through standardization, the data can be scaled to a range with a mean of 0 and a standard deviation of 1.

[0043] For example, the original value at 500nm is 1.45, and it may become 0.2 after standardization. This method can eliminate the influence of dimensions and facilitate subsequent analysis. Based on the standardized feature data set, calculating the contribution rate of the eigenvectors through variance calculation is the screening basis.

[0044] In one embodiment, the calculation results may show that the contribution rate of the first eigenvector is 80%, the second is 15%, and the third is only 3%.

[0045] Preferably, the contribution rate threshold is set at 5%, then the third eigenvector is eliminated to reduce redundant information and retain the core features. For the refined feature dataset, the clustering algorithm can divide the wavelength range to reveal the data structure.

[0046] For example, using the K-means algorithm, the range from 200nm to 1000nm is divided into three groups: 200 - 400nm, 400 - 700nm, and 700 - 1000nm.

[0047] Specifically, after running the algorithm, the first group may concentrate around 1.35, the second group around 1.50, and the third group around 1.45. This grouping reflects the differences in wavelength characteristics and helps with targeted research. By grouping according to the wavelength range, the average refractive index within each group is calculated to further refine the information.

[0048] In one possible implementation, the average value for 200 - 400nm is 1.34, for 400 - 700nm is 1.48, and for 700 - 1000nm is 1.46, forming an average value table.

[0049] It should be noted that this average value table intuitively shows the characteristics of each section and is convenient for comparison. Based on the average value table of the wavelength range, the key wavelength range reflecting the material quality is the ultimate goal. In this embodiment, the finally determined key wavelength range is the second data matrix, which includes the target incident light within the key wavelength range, and also includes the second refractive index of the nanomaterial with the second uniformity value, second purity value, and second refractive index under the irradiation of the target incident light.

[0050] For example, if the average value of 400 - 700nm is 1.48 and the fluctuation is small, it may indicate that the optical performance in this range is stable and suitable for specific applications.

[0051] Exemplarily, from multiple aspects, if the average value of this range is consistent in different samples and significantly different from other ranges, then it can be confirmed as the key range. In another embodiment, if the average value of 200 - 400nm is low and the fluctuation is large, it may indicate material defects.

[0052] Preferably, combined with the actual needs, focus on the stable range to generate the final feature dataset. This method can effectively support the optimized design of materials.

[0053] In this embodiment, from dimensionality reduction to feature extraction, and then to grouped analysis, the layers progress step by step to ensure the refinement and practicality of the data. Specifically, principal component analysis reduces the dimensions, standardization unifies the scale, and clustering and mean calculation highlight the characteristics. The logical chain of this embodiment not only simplifies complex data but also provides a reliable basis for material research.

[0054] Step 103: Based on the support vector regression algorithm, establish a quality detection model representing the non-linear relationship between the second data matrix, the second consistency degree value, the second uniformity value, and the second purity value. When inputting the target refractive index of the nanomaterial into the quality detection model, the quality detection model outputs the target consistency degree value, the target uniformity value, and the target purity value of the nanomaterial.

[0055] The support vector regression algorithm can map the input data to a high-dimensional space and find an optimal hyperplane in the high-dimensional space so that the deviation of all sample points from the hyperplane does not exceed a preset threshold. Based on this principle, this embodiment establishes a quality detection model that can combine the refractive index of the nanomaterial under different incident light irradiations, as well as the consistency degree value, a uniformity value, and a purity value of the nanomaterial to form a quality detection model. The generated quality detection model can directly perform quality detection on the nanomaterial based on the refractive index, providing strong support for the quality control and performance optimization of the nanomaterial.

[0056] In some embodiments, based on the support vector regression algorithm, establishing a quality detection model representing the non-linear relationship between the second data matrix, the second consistency degree value, the second uniformity value, and the second purity value includes:

[0057] Based on the support vector regression algorithm, establish an initial model representing the second data matrix, the second consistency degree value, the second uniformity value, and the second purity value; use the test data set to iteratively modify the model parameters in the initial model until the initial model converges to obtain the target quality detection model. The model parameters include the kernel function value and the penalty coefficient value. The test data set contains part of the data in the second data matrix and the corresponding second consistency degree value, second uniformity value, and second purity value; use the target quality detection model to identify and delete the abnormal data in the second data matrix to obtain the third data matrix and the corresponding third consistency degree value, third uniformity value, and third purity value; based on the target quality detection model, establish a quality detection model representing the non-linear relationship between the third data matrix, the third consistency degree value, the third uniformity value, and the third purity value.

[0058] By performing a modeling operation on the dimension-reduced feature data set, an initial model is obtained. For the initial model, the cross-validation method is used to evaluate the model stability to obtain a verification result. According to the verification result, if the model stability is lower than the preset threshold, the support vector parameters are adjusted to obtain an optimized initial model as the quality detection model.

[0059] Specifically, performing a modeling operation on the dimension-reduced first data matrix set through the support vector regression algorithm is an important means of transforming the extracted features into a quality detection model. The core of support vector regression lies in finding a hyperplane that can maximize the margin and minimize the prediction error.

[0060] The dimension-reduced second data matrix contains the refractive indices in multiple wavelength ranges. For example, the average refractive index of the incident light corresponding to 200 - 400 nm is 1.34, and the average refractive index of the incident light corresponding to 400 - 700 nm is 1.48. These can be used as the input variables of the initial model to predict the consistency value, uniformity value, and purity value of the nanomaterials.

[0061] In a possible implementation, the algorithm in the initial model will construct a regression function based on the second data matrix and the corresponding second uniformity value, second purity value, and second refractive index, attempting to fit the data distribution.

[0062] Specifically, the parameters of support vector regression, such as the kernel function and penalty coefficient, directly affect the model performance.

[0063] In one embodiment, if the algorithm selects a linear kernel function, the model may tend to capture the linear trend of the data, such as the simple relationship between the wavelength range mean and the quality parameters. If the algorithm switches to a radial basis kernel function, it can better fit the non-linear characteristics, such as the complex influence of the mean fluctuation in certain wavelength ranges on the quality.

[0064] Preferably, the algorithm can be made to start with a linear kernel first. If a large prediction deviation is found, it can be adjusted to a non-linear kernel. This flexibility ensures that the model can adapt to different data characteristics and improve the prediction ability. For the regression model, using the cross-validation method to evaluate the stability is a key step to ensure reliable results.

[0065] It can be understood that cross-validation can be carried out by dividing the second data matrix into a training set and a test set, and verifying the model performance through multiple iterations.

[0066] For example, the refractive index feature data set is divided into 5 parts, and 4 parts are used for training and 1 part for testing in turn, and finally the average error index is obtained.

[0067] In a possible implementation, assume that the mean squared errors of each test are 0.05, 0.06, 0.04, 0.05, and 0.07 respectively, and the average value is 0.054. This quantified result intuitively reflects the stability of the model and is convenient for subsequent optimization. According to the verification result, if the model stability is lower than the preset threshold, adjusting the support vector parameters is a necessary step.

[0068] Exemplarily, assume that the threshold is the mean squared error of 0.05, and the actual value is 0.054, indicating that the model is slightly unstable.

[0069] In some embodiments, by using a test data set, iteratively modify the model parameters in the initial model until the initial model converges to obtain a target quality detection model, including:

[0070] Use the cross-validation method to determine the model validation result of the initial model through the test data set;

[0071] If the model validation result is lower than the first threshold, determine the model parameters of the support vector regression algorithm, where the first threshold is used to represent the stability threshold of the initial model;

[0072] Iteratively increase the penalty coefficient value and / or reduce the width parameter value of the kernel function until the initial model converges to obtain a target quality detection model, where the width parameter value is used as the kernel function value.

[0073] Specifically, the penalty coefficient can be increased, adjusted from the default 1 to 10, to strengthen the constraint on errors; or the width parameter of the kernel function can be reduced, adjusted from 0.1 to 0.01, to improve the model's sensitivity to local features.

[0074] In one embodiment, the adjusted error may drop to 0.045, meeting the expected requirements. This optimization process not only improves the prediction accuracy but also enhances the model's generalization ability for unknown data.

[0075] It should be noted that the risk of overfitting needs to be considered when adjusting the model parameters.

[0076] For example, if the penalty coefficient is too high, the model may fit the training data too well, resulting in an increase in the test error instead.

[0077] Preferably, the grid search method can be combined to systematically test parameter combinations, such as penalty coefficient values of 1, 5, 10 and kernel width values of 0.01, 0.05, 0.1, and finally find the best combination. This method ensures that the optimized regression model is both stable and practical through multi-dimensional verification, providing strong support for subsequent material quality analysis.

[0078] In some embodiments, by using the target quality detection model, identify and delete the abnormal data in the second data matrix to obtain a third data matrix, and the corresponding third consistency degree value, third uniformity value, and third purity value of the third data matrix, including:

[0079] Input the second data matrix into the target quality detection model to obtain the prediction results output by the target quality detection model, where the prediction results include the first prediction value of the consistency degree, the first prediction value of the uniformity, and the first prediction value of the purity;

[0080] Using the result difference between the prediction result and the actual result, determine the first abnormal data set in the second data matrix, where the actual result represents the second consistency degree value, the second uniformity value, and the second purity value corresponding to the second data matrix respectively;

[0081] Using the K-means clustering algorithm, classify the first abnormal data set to obtain a second abnormal data set, and the abnormal risk degree of the second abnormal data set is higher than that of the first abnormal data set;

[0082] Delete the second abnormal data set in the second data matrix to obtain a third data matrix, and the third consistency degree value, the third uniformity value, and the third purity value corresponding to the third data matrix.

[0083] Perform a prediction operation on the second data matrix through the target quality detection model to obtain a prediction result of the material quality. Calculate the deviation between the prediction result and the actual result to obtain the deviation value of each sample in the second data matrix. If the deviation value exceeds the preset threshold, mark the corresponding sample as abnormal data to obtain a marking result. Use cluster analysis to group the abnormal data in the marking result to obtain the distribution characteristics of the abnormal data. Sort the abnormal data according to the distribution characteristics to determine the high-risk abnormal sample set as the second abnormal data set, and delete the second abnormal data set.

[0084] Specifically, input the second data matrix into the target quality detection model, so that the target quality detection model performs a prediction operation on the second data matrix, and obtains and outputs a prediction result.

[0085] For example, assume that the second data matrix contains the refractive indices of nanomaterials under the irradiation of the same target incident light. After prediction by the target quality detection, a set of predicted quality parameter values are obtained. For example, the prediction result of nanomaterial sample A is 85, and the actual result of nanomaterial sample B is 90. Calculating the deviation between the prediction result and the actual result can intuitively reflect the accuracy of the model.

[0086] Exemplarily, if the measured value of nanomaterial sample A is 87, the deviation is 2; the actual result of nanomaterial sample B is 88, and the deviation is -2. The key to deviation calculation is to quantify the gap between prediction and actual, providing a basis for subsequent anomaly identification.

[0087] In a possible implementation, if the preset deviation threshold is 3, the data in the second data matrix whose deviation from the target quality detection model exceeds the threshold is marked as abnormal data.

[0088] Specifically, assume that the predicted value of nanomaterial sample C is 75, the measured value is 80, and the deviation is 5, exceeding the threshold, then it is marked as abnormal. This marking method can quickly screen out potential problem samples.

[0089] It should be noted that the threshold setting needs to be flexibly adjusted according to the material quality requirements. For example, it can be set to 2 for high-precision materials and relaxed to 5 for general materials to balance false positives and false negatives. Using cluster analysis to group abnormal data can reveal its distribution characteristics.

[0090] For example, when inputting abnormal samples into the clustering algorithm, it may be found that the deviation of one group of samples is concentrated between 4 and 6, and the other group is concentrated between -5 and -3.

[0091] Preferably, if K-means clustering is used, the number of clusters can be preset to 3, and observe whether the abnormal samples are related to specific characteristics, such as refractive index fluctuations or measurement conditions.

[0092] It can be understood that this grouping helps to mine the potential laws of abnormalities from the data distribution. Sorting the abnormal data according to the distribution characteristics is a key step in determining high-risk samples.

[0093] In one embodiment, if the deviation of a certain cluster of samples is large and the quantity is concentrated, such as the deviation of 10 samples all exceeding 5, then the priority is higher and it is classified into the high-risk abnormal sample set.

[0094] Specifically, it can be sorted according to the mean deviation. The higher the mean, the greater the risk.

[0095] For example, if the mean of cluster A is 5.5 and the mean of cluster B is 3.2, then cluster A has a higher priority. This sorting method provides a clear direction for subsequent verification. By performing secondary verification on the high-risk abnormal sample set through logistic regression, the reliability of abnormal identification can be improved.

[0096] Exemplarily, assume that the high-risk set contains 20 samples. Logistic regression takes the deviation characteristics as input and outputs the abnormal probability. If the probability of sample D is 0.9 and the probability of sample E is 0.6, and the probability threshold is set to 0.8, then sample D is confirmed as the final abnormality.

[0097] In one possible implementation, additional features can be introduced, such as sample collection time or environmental variables, to enrich the model judgment basis.

[0098] Preferably, logistic regression can also adjust the weights in combination with prior knowledge. For example, if it is known that a certain wavelength range has a greater impact on quality, its weight can be increased. This flexibility makes the verification process more in line with actual needs. The final abnormal identification result can not only be used for quality control, but also provide data support for process improvement.

[0099] In some embodiments, deleting the second abnormal data set in the second data matrix to obtain a third data matrix includes:

[0100] Using the random forest algorithm, classifying the second abnormal data set based on the wavelength quantity value of the target incident light to obtain at least one second abnormal sub-data set;

[0101] Determine the distribution characteristic value of the second abnormal sub - dataset based on the time - series characteristics of the second abnormal dataset;

[0102] When the distribution characteristic of the second abnormal sub - dataset is greater than the preset distribution deviation value, perform fluctuation analysis on the second abnormal sub - dataset to obtain a fluctuation analysis result. The fluctuation analysis result is used to represent the correlation between the second abnormal sub - dataset and the abnormal consistency degree value, abnormal uniformity value, and abnormal purity value. The abnormal consistency degree value, abnormal uniformity value, and abnormal purity value respectively belong to the second consistency degree value, the second uniformity value, and the second purity value;

[0103] If the fluctuation analysis result is higher than the fluctuation threshold, delete the second abnormal sub - dataset to obtain a third data matrix.

[0104] The random forest algorithm is an ensemble learning method based on decision trees. By combining the prediction results of multiple decision trees, it improves the accuracy and robustness of the model. Its core idea combines random sampling and the randomness of the splitting characteristics of decision trees, effectively reducing the risk of overfitting.

[0105] Perform a classification operation on the second abnormal dataset through the random forest algorithm to obtain the classified second abnormal sub - dataset. Extract time - series characteristics for the second abnormal sub - dataset to obtain a feature change trend description matrix. Calculate the statistical distribution between samples based on the feature change trend description matrix to determine the distribution deviation degree index. If the distribution deviation degree index exceeds the preset threshold, then label the second abnormal sub - dataset through pattern recognition technology to obtain a labeled high - risk sample set. Construct a parameter fluctuation analysis model using the labeled high - risk sample set to judge the correlation with material quality and obtain a fluctuation impact assessment result. Screen the abnormal sample subset through the fluctuation impact assessment result to obtain the finally confirmed abnormal data set, and delete this abnormal data set.

[0106] Specifically, performing a classification operation on the abnormal data set through the random forest algorithm can effectively separate the abnormal sample subset. Exemplarily, assume there is a data set containing 50 abnormal samples, and each sample has a refractive index value of a single wavelength. The random forest divides these 50 samples into two categories by constructing multiple decision trees: 30 are classified into the mild abnormal subset, and 20 are classified into the severe abnormal subset. This classification method utilizes the comprehensive judgment of multiple features and is suitable for dealing with the diversity of abnormal data. Extract time - series characteristics for the classified abnormal sample subset to obtain a feature change trend description matrix.

[0107] Specifically, assume that 20 severe abnormal samples have collected refractive index data at 5 consecutive time points. In one possible implementation, the extracted features include the refractive index mean and change rate of each sample.

[0108] For example, the refractive index sequence of sample X is 1.50, 1.52, 1.55, 1.53, 1.51, with an average value of 1.52 and a change rate of 0.03; the sequence of sample Y is 1.60, 1.62, 1.65, 1.63, 1.61, with an average value of 1.62 and a change rate of 0.04. The characteristic change trend description matrix is thus formed, including two columns of the average values and change rates of 20 samples, which helps to reveal the law of the evolution of anomalies over time. Calculate the statistical distribution among samples according to the characteristic change trend description matrix, and determine the distribution deviation degree index.

[0109] In one embodiment, it can be calculated through the distribution ranges of the sample average value and the change rate.

[0110] For example, the average value range of 20 samples is between 1.50 and 1.65, and the change rate range is between 0.02 and 0.05. If the average value distribution of normal samples should be concentrated between 1.50 and 1.55, and the change rate is between 0.01 and 0.03, then the current distribution deviation degree index can be set to 0.15.

[0111] Preferably, the threshold is set to 0.10. When the index exceeds this value, it indicates that the distribution of abnormal samples significantly deviates from the normal range, providing a basis for subsequent processing. If the distribution deviation degree index exceeds the preset threshold, label the subset of abnormal samples through pattern recognition technology to obtain a high-risk sample set.

[0112] For example, based on the distribution characteristics of the average value and the change rate, the pattern recognition technology identifies that the patterns of 10 samples are significantly different from those of normal samples, such as the average value continuously being higher than 1.60 and the change rate exceeding 0.04, and these samples are labeled as high-risk.

[0113] It should be noted that this labeling method can focus on the core characteristics of anomalies and improve the pertinence of screening. Construct a parameter fluctuation analysis model using the labeled high-risk sample set to judge the correlation of material quality and obtain the fluctuation impact evaluation result.

[0114] In one possible implementation, for 10 high-risk samples, analyze the frequency and amplitude of the refractive index fluctuation.

[0115] For example, the refractive index of sample Z fluctuates frequently between 1.60 and 1.65, which may imply the instability of the internal structure of the material. The fluctuation impact evaluation result shows that the fluctuations of 8 samples are related to material quality problems, providing clues for subsequent verification. Screen the subset of abnormal samples through the fluctuation impact evaluation result to obtain the finally confirmed abnormal data set.

[0116] It is understandable that 8 samples with significant fluctuations are selected from 10 high-risk samples as the final abnormal data set, and the other 2 may be caused by external interference. This method improves the accuracy of anomaly recognition and provides more reliable data support for material quality analysis.

[0117] After determining that the second abnormal sub-data set is abnormal data, the second abnormal sub-data set is removed, and the data set is updated in the second data matrix to obtain a third data matrix.

[0118] In some embodiments, based on the target quality detection model, a quality detection model establishing a non-linear relationship between the third data matrix, the consistency degree value, the uniformity value, and the purity value includes:

[0119] Based on the target quality detection model, establish a non-linear relationship between the third data matrix, the consistency degree value, the uniformity value, and the purity value;

[0120] Determine the distribution ranges of the consistency degree value, the uniformity value, and the purity value;

[0121] Within the distribution range, use the interpolation method to update the third data matrix to obtain a fourth data matrix;

[0122] Establish a quality detection model establishing a non-linear relationship between the fourth data matrix, the consistency degree value, the uniformity value, and the purity value.

[0123] According to the target quality detection model, obtain the correlation characteristics between the material quality and the refractive index data, and judge the distribution range of the material quality parameters. For the material quality parameters within the distribution range, use the interpolation method to process the refractive index data to obtain a smoothed data set. Update the mapping relationship through the smoothed data set to obtain an adjusted evaluation model. Extract key features from the adjusted evaluation model to obtain the final quality evaluation basis.

[0124] Specifically, constructing a data matrix through refractive index data is a basic link for the application of the support vector regression algorithm.

[0125] Exemplarily, assume that 80 refractive index samples are collected in the experiment, the wavelength is fixed at 550 nanometers, and the values are distributed between 1.50 and 1.60. A matrix with 80 rows and 1 column can be designed, and each row records a refractive index value, such as 1.52, 1.57, etc. This matrix structure is convenient for inputting into the algorithm for subsequent processing.

[0126] In a possible implementation manner, the support vector regression algorithm captures the non-linear relationship between data by training this matrix. Assume that the radial basis function is selected as the kernel function during training, and the model can initially fit the change trend of the refractive index to generate an initial training model. Extracting the multivariate analysis results from the initial training model is the key to determining the mapping relationship.

[0127] Specifically, the model output may show the correlation between the refractive index and material quality parameters such as density or hardness.

[0128] For example, when the refractive index increases from 1.50 to 1.60, the density may show an increasing trend.

[0129] It should be noted that the accuracy of this mapping relationship depends on the representativeness of the training data. If the mapping relationship deviates from the preset threshold, such as the correlation coefficient being lower than 0.8, the model needs to be adjusted.

[0130] Preferably, noise samples can be removed through statistical analysis. For example, samples that deviate too far from the mean are marked as abnormal, such as 1.49 or 1.61, and then retrained to obtain an optimized model. Obtaining the associated features based on the optimized model is an important step in quality assessment.

[0131] In one embodiment, the model may reveal that the best hardness value corresponds to around a refractive index of 1.55, and the hardness decreases when deviating from this value.

[0132] For example, the samples with refractive indices of 1.53 and 1.57 have higher hardness, while those with 1.50 and 1.60 are lower. This feature helps to judge the distribution range of the quality parameters. For the quality parameters within the distribution range, interpolation can smooth the data.

[0133] Exemplarily, if the original data points are sparse, linear interpolation can be used to generate intermediate values such as 1.55 between 1.54 and 1.56 to obtain a continuous data set. This smoothing process improves the uniformity of the data. Updating the mapping relationship based on the smoothed data set is the core of model adjustment.

[0134] It can be understood that the updated evaluation model can better reflect the true relationship between the refractive index and quality.

[0135] For example, after interpolation, the number of data points increases, and the model may show that the hardness peak is more concentrated in the range of 1.54 to 1.56.

[0136] In one possible implementation, key features are extracted from the adjusted model, such as the correlation between the refractive index change rate and quality stability.

[0137] Specifically, if the refractive index fluctuation is less than 0.02, the quality parameters may be more stable. This feature provides a basis for the final quality assessment.

[0138] For example, based on this, it can be judged that the material quality with the refractive index controlled between 1.54 and 1.56 is better. This method ensures the reliability of the assessment through multi-step optimization and provides practical support for material screening.

[0139] In some embodiments, after establishing the quality detection model, the method further includes:

[0140] After inputting the target refractive index of the nanomaterial to be tested into the quality inspection model, a quality parameter interval output by the quality inspection model is obtained. The quality parameter interval includes the variation intervals of the second predicted value of the degree of consistency, the second predicted value of the uniformity, and the second predicted value of the purity.

[0141] According to the intermediate value of the variation trend of the quality parameter interval, the parameter values in the quality parameter interval are updated, and the updated quality parameter interval is mapped to the refractive index to be tested of the nanomaterial to be tested.

[0142] Based on the mapping relationship between the quality parameter interval and the refractive index to be tested of the nanomaterial to be tested, the quality inspection model outputs the target degree of consistency value, the target uniformity value, and the target purity value corresponding to the target refractive index of the nanomaterial.

[0143] Process the nanomaterial data through the adjusted quality inspection model to obtain the preliminary distribution of the quality parameters. Extract the multi-wavelength characteristics from the preliminary distribution, and use data mining techniques to determine the variation trend of the uniformity index. Taking the purity value as an example, fuse the structural consistency characteristics according to the variation trend to obtain the intermediate result of the purity value evaluation. If the intermediate result deviates from the preset threshold, adjust the quantization index by interpolation method to obtain the smoothed data set. Update the fusion technology through the smoothed data set to determine the mapping relationship of the output content. Extract the characterization result according to the mapping relationship to obtain the final purity value distribution.

[0144] Specifically, when processing the target refractive index of the nanomaterial through the quality inspection model, the core is to extract the preliminary distribution of the quality parameters from the complex data.

[0145] Exemplarily, assume that 100 refractive index samples of nanomaterials are collected, and the wavelength range covers 400 to 700 nanometers. The preliminary distribution may show that the quality parameters such as particle size fluctuate between 10 and 50 nanometers. In one possible implementation, after the model inputs these data, a size distribution diagram is generated through multivariate analysis to intuitively reflect the central tendency of the quality parameters.

[0146] In one embodiment, if the refractive index is stable at about 1.50 at 550 nanometers and the lattice regularity is confirmed by combining X-ray diffraction data, it can be inferred that the purity is high. On the contrary, if the refractive index jumps significantly when the wavelength changes, such as rising from 1.45 to 1.60, there may be impurities. This fusion method helps to improve the comprehensiveness of the evaluation. If the intermediate result deviates from the preset threshold, for example, the purity evaluation value is lower than 90%.

[0147] Exemplarily, when the refractive index data points are sparse between 400 and 550 nanometers, intermediate values such as 1.48 can be inserted to generate a smooth curve. This adjustment makes the data more continuous and facilitates subsequent analysis. When updating the fusion technology with the smoothed data set, one possible implementation is to introduce the weighted average method. Assuming the weight of 400 nanometers is 0.3 and that of 700 nanometers is 0.7, recalculate the mapping relationship to highlight the influence of key wavelengths on purity.

[0148] For example, it may be found after adjustment that the data at 700 nanometers contributes more to purity. This update ensures that the output mapping is closer to reality. When extracting the characterization results according to the mapping relationship.

[0149] Specifically, eigenvalue can be obtained from the model, such as the correlation between the refractive index change rate and purity.

[0150] In one embodiment, if the change rate is less than 0.01, the purity is stable above 95%, while when the change rate rises to 0.05, the purity drops to 85%. This characterization provides a basis for the final distribution. When obtaining the final distribution of quality parameters.

[0151] For example, the size distribution may be concentrated between 20 and 30 nanometers, and the proportion of samples with a purity higher than 93% reaches 80%.

[0152] It can be understood that this distribution provides intuitive support for screening high-quality nanomaterials and at the same time optimizes the subsequent process design. This method ensures the reliability of the results through multi-faceted verification.

[0153] As Figure 2 shown, this embodiment discloses a structural diagram of a quality detection device for nanomaterials, including:

[0154] An acquisition module 21, configured to acquire the first refractive index of the nanomaterial when irradiated by target incident light at different wavelengths. The combination of the acquired first refractive index and the wavelength of the target incident light forms a first data matrix. The nanomaterial includes at least one consistency degree value, at least one uniformity value, and at least one purity value. In the first data matrix, the first refractive index and the wavelength of the target incident light correspond one by one;

[0155] A selection module 22, configured to reduce the dimension of the first data matrix by using the principal component analysis method, select a target wavelength range in the dimension-reduced first data matrix, and use the wavelength of the target incident light included in the target wavelength range and the first refractive index corresponding to the wavelength of the target incident light to form a second data matrix. The information quantity value in the second data matrix is the highest. The first uniformity value, the first purity value, and the first refractive index corresponding to the second data matrix are used as the second uniformity value, the second purity value, and the second refractive index respectively;

[0156] A building module 23 is configured to establish a quality detection model based on a support vector regression algorithm, which represents the non-linear relationship between the second data matrix and the second consistency degree value, the second uniformity value, and the second purity value. When the target refractive index of the nanomaterial is input into the quality detection model, the quality detection model outputs the target consistency degree value, the target uniformity value, and the target purity value of the nanomaterial.

[0157] Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, a USB flash drive, a mobile hard disk, etc.), and includes several instructions for causing a computer device (such as a personal computer, a server, or a network device, etc.) to execute the methods of various implementation scenarios of the present application.

[0158] Optionally, the above-mentioned entity device may further include a user interface, a network interface, a camera, a radio frequency (RF) circuit, sensors, an audio circuit, a WI-FI module, and so on. The user interface may include a display screen (Display), an input unit such as a keyboard (Keyboard), etc. Optionally, the user interface may further include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface), etc.

[0159] Those skilled in the art can understand that the above-mentioned entity device structure provided in this embodiment does not limit the entity device, and it may include more or fewer components, or combine certain components, or have different component arrangements.

[0160] Based on the above method as Figure 1 shown, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the method corresponding to any embodiment. The storage medium may further include an operating system and a network communication module. The operating system is a program for managing the hardware and software resources of the above-mentioned entity device, and supports the operation of the information processing program and other software and / or programs. The network communication module is used to implement the communication between the components inside the storage medium, and the communication between other hardware and software in the information processing entity device.

[0161] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform, or can also be implemented by hardware. By applying the solution of this embodiment, compared with the current existing technologies, when the present embodiment obtains the first refractive index of the nanomaterial under the irradiation of target incident light at different wavelengths, the combination of the obtained first refractive index and the wavelength of the target incident light is the first data matrix. The nanomaterial includes at least one first degree of consistency value, at least one first uniformity value, and at least one first purity value. In the first data matrix, the first refractive index and the wavelength of the target incident light correspond one by one. Using the principal component analysis method to reduce the dimension of the first data matrix, select the target wavelength range in the dimension-reduced first data matrix, and use the wavelength of the target incident light included in the target wavelength range and the first refractive index corresponding to the wavelength of the target incident light to form the second data matrix, and the information content value of the second data matrix is the highest. The first uniformity value, the first purity value, and the first refractive index corresponding to the second data matrix are respectively used as the second uniformity value, the second purity value, and the second refractive index. Based on the support vector regression algorithm, a quality detection model representing the non-linear relationship between the second data matrix, the second degree of consistency value, the second uniformity value, and the second purity value is established. When the target refractive index of the nanomaterial is input into the quality detection model, the quality detection model outputs the target degree of consistency value, the target uniformity value, and the target purity value of the nanomaterial. The present application combines the first refractive index of the nanomaterial under the irradiation of different incident lights, as well as the degree of consistency value, a uniformity value, and a purity value of the nanomaterial to construct a quality detection model. The constructed quality detection model can directly perform quality detection on the nanomaterial based on the refractive index, providing strong support for the quality control and performance optimization of the nanomaterial.

[0162] It should be noted that in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article, or device. Without further limitation, an element defined by the statement "comprising one......" does not exclude the existence of additional identical elements in the process, method, article, or device comprising the element.

[0163] The above are only specific embodiments of the present application, enabling those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments described herein, but rather will conform to the broadest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for quality inspection of a nanomaterial, characterized in that, Including: Obtaining a first refractive index of a nanomaterial when irradiated with target incident light at different wavelengths, and combining the obtained first refractive index and the wavelengths of the target incident light into a first data matrix. The nanomaterial includes at least one first degree of consistency value, at least one first uniformity value, and at least one first purity value, and in the first data matrix, the first refractive index and the wavelengths of the target incident light correspond one by one; Using the principal component analysis method to reduce the dimension of the first data matrix, selecting a target wavelength range in the reduced first data matrix, and using the wavelengths of the target incident light included in the target wavelength range and the first refractive indices corresponding to the wavelengths of the target incident light to form a second data matrix, where the information content value of the second data matrix is the highest, and the first uniformity value, the first purity value, and the first refractive index corresponding to the second data matrix are used as the second uniformity value, the second purity value, and the second refractive index respectively; Based on the support vector regression algorithm, establishing a quality detection model representing the non-linear relationship between the second data matrix, the second degree of consistency value, the second uniformity value, and the second purity value. When inputting the target refractive index of the nanomaterial into the quality detection model, the quality detection model outputs the target degree of consistency value, target uniformity value, and target purity value of the nanomaterial.

2. The method according to claim 1, wherein Based on the support vector regression algorithm, establishing a quality detection model representing the non-linear relationship between the second data matrix, the second degree of consistency value, the second uniformity value, and the second purity value, including: Based on the support vector regression algorithm, establishing an initial model representing the non-linear relationship between the second data matrix, the second degree of consistency value, the second uniformity value, and the second purity value; Using a test data set to iteratively modify the model parameters in the initial model until the initial model converges to obtain a target quality detection model. The model parameters include kernel function values and penalty coefficient values. The test data set includes some data in the second data matrix, and the second degree of consistency value, the second uniformity value, and the second purity value corresponding to the partial data; Using the target quality detection model to identify and delete abnormal data in the second data matrix to obtain a third data matrix, and a third degree of consistency value, a third uniformity value, and a third purity value corresponding to the third data matrix; Based on the target quality detection model, establishing a quality detection model representing the non-linear relationship between the third data matrix, the third degree of consistency value, the third uniformity value, and the third purity value.

3. The method according to claim 2, characterized in that The step of using a test data set to iteratively modify the model parameters in the initial model until the initial model converges to obtain a target quality detection model includes: Using the cross-validation method to determine the model validation result of the initial model through the test data set; If the model validation result is lower than a first threshold, determining the model parameters supporting the vector regression algorithm, where the first threshold is used to represent the stability threshold of the initial model; Iteratively increase the penalty coefficient value and / or reduce the width parameter value of the kernel function until the initial model converges to obtain a target quality detection model, where the width parameter value serves as the kernel function value.

4. The method according to claim 2, wherein The step of using the target quality detection model to identify and delete the abnormal data in the second data matrix to obtain a third data matrix, and the corresponding third consistency degree value, third uniformity value, and third purity value, includes: Input the second data matrix into the target quality detection model to obtain the prediction results output by the target quality detection model, where the prediction results include the first predicted value of the consistency degree, the first predicted value of the uniformity, and the first predicted value of the purity; Use the result difference between the prediction result and the actual result to determine the first abnormal data set in the second data matrix, where the actual result represents the second consistency degree value, the second uniformity value, and the second purity value corresponding to the second data matrix respectively; Use the K-means clustering algorithm to classify the first abnormal data set to obtain a second abnormal data set, where the abnormal risk degree of the second abnormal data set is higher than that of the first abnormal data set; Delete the second abnormal data set in the second data matrix to obtain a third data matrix, and the corresponding third consistency degree value, third uniformity value, and third purity value.

5. The method according to claim 4, characterized in that, The step of deleting the second abnormal data set in the second data matrix to obtain a third data matrix includes: Use the random forest algorithm to classify the second abnormal data set based on the wavelength quantity value of the target incident light to obtain at least one second abnormal sub-data set; Based on the time series feature of the second abnormal data set, determine the distribution feature value of the second abnormal sub-data set; When the distribution feature of the second abnormal sub-data set is greater than the preset distribution deviation value, perform fluctuation analysis on the second abnormal sub-data set to obtain a fluctuation analysis result, where the fluctuation analysis result is used to represent the correlation between the second abnormal sub-data set and the abnormal consistency degree value, abnormal uniformity value, and abnormal purity value, and the abnormal consistency degree value, abnormal uniformity value, and abnormal purity value respectively belong to the second consistency degree value, the second uniformity value, and the second purity value; If the fluctuation analysis result is higher than the fluctuation threshold, delete the second abnormal sub-data set to obtain a third data matrix.

6. The method according to claim 5, characterized in that, The step of establishing a quality detection model based on the target quality detection model to represent the non-linear relationship between the third data matrix and the second consistency degree value, the second uniformity value, and the second purity value includes: Determine the distribution range of the third consistency degree value, the third uniformity value, and the third purity value; Within the distribution range, use the interpolation method to update the third data matrix to obtain a fourth data matrix, and update the third consistency degree value, the third uniformity value, and the third purity value to obtain the corresponding fourth consistency degree value, fourth uniformity value, and fourth purity value of the fourth data matrix; Establish a quality detection model that has a non-linear relationship between the fourth data matrix, the fourth consistency degree value, the fourth uniformity value, and the fourth purity value.

7. The method according to any one of claims 1 to 6, characterized in that After establishing the quality detection model, the method further includes: After inputting the target refractive index of the nanomaterial to be measured into the quality detection model, obtain the quality parameter interval output by the quality detection model. The quality parameter interval includes the change intervals of the second predicted value of the consistency degree, the second predicted value of the uniformity, and the second predicted value of the purity; According to the median value of the change trend of the quality parameter interval, update the parameter values in the quality parameter interval. The updated quality parameter interval is mapped to the refractive index to be measured of the nanomaterial to be measured; Based on the mapping relationship between the quality parameter interval and the refractive index to be measured of the nanomaterial to be measured, the quality detection model outputs the target consistency degree value, the target uniformity value, and the target purity value corresponding to the target refractive index of the nanomaterial to be measured.

8. A quality inspection device for a nanomaterial, characterized in that, Includes: An acquisition module for acquiring the first refractive index of the nanomaterial when irradiated by the target incident light at different wavelengths. The combination of the acquired first refractive index and the wavelength of the target incident light is the first data matrix. The nanomaterial includes at least one first consistency degree value, at least one first uniformity value, and at least one first purity value. In the first data matrix, the first refractive index and the wavelength of the target incident light correspond one by one; A selection module for using the principal component analysis method to reduce the dimension of the first data matrix, selecting the target wavelength interval in the reduced first data matrix, and using the wavelength of the target incident light included in the target wavelength interval and the first refractive index corresponding to the wavelength of the target incident light to form the second data matrix. The information content value of the second data matrix is the highest. The first uniformity value, the first purity value, and the first refractive index corresponding to the second data matrix are used as the second uniformity value, the second purity value, and the second refractive index respectively; A building module for establishing a quality detection model that represents the non-linear relationship between the second data matrix, the second consistency degree value, the second uniformity value, and the second purity value based on the support vector regression algorithm. When the target refractive index of the nanomaterial is input into the quality detection model, the quality detection model outputs the target consistency degree value, the target uniformity value, and the target purity value of the nanomaterial.

9. An electronic device, characterized in that, Includes: At least one processor; And A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method according to any one of claims 1-7.

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

Images